From f9b462a37e9d941b7349e2e100cc3e6e5c076369 Mon Sep 17 00:00:00 2001 From: YurenHao0426 Date: Wed, 22 Jul 2026 03:31:14 -0500 Subject: results: retain failed useful-depth local panel --- results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s4.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s0.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s1.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s2.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s3.json | 1 + results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s4.json | 1 + 120 files changed, 120 insertions(+) create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s4.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s0.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s1.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s2.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s3.json create mode 100644 results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s4.json diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s0.json b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s0.json new file mode 100644 index 0000000..f171995 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d1_t0_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7099499106407166}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.7525, "eval_loss": 0.5949873352050781, "wall_s": 4.132476091384888, "val_acc": 0.7525, "val_loss": 0.5949873352050781}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 4.1033244132995605, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 4.1033244132995605, "evaluation_wall_s": 0.023546695709228516}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17273856, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s1.json b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s1.json new file mode 100644 index 0000000..35a8894 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d1_t0_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6922825574874878}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.741, "eval_loss": 0.5997843322753906, "wall_s": 4.179248332977295, "val_acc": 0.741, "val_loss": 0.5997843322753906}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 4.151807546615601, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 4.151807546615601, "evaluation_wall_s": 0.02183365821838379}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17273856, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s2.json b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s2.json new file mode 100644 index 0000000..dac9c74 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d1_t0_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7055633664131165}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.7155, "eval_loss": 0.6046191711425781, "wall_s": 3.927260637283325, "val_acc": 0.7155, "val_loss": 0.6046191711425781}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 3.903226613998413, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 3.903226613998413, "evaluation_wall_s": 0.01878952980041504}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17273856, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s3.json b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s3.json new file mode 100644 index 0000000..224b35b --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d1_t0_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7144617438316345}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.724, "eval_loss": 0.603665771484375, "wall_s": 4.078364372253418, "val_acc": 0.724, "val_loss": 0.603665771484375}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 4.051671266555786, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 4.051671266555786, "evaluation_wall_s": 0.021528005599975586}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17273856, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s4.json b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s4.json new file mode 100644 index 0000000..c230465 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t0_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d1_t0_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7867427468299866}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.69, "eval_loss": 0.5876565246582032, "wall_s": 4.2396814823150635, "val_acc": 0.69, "val_loss": 0.5876565246582032}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 4.207880973815918, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 4.207880973815918, "evaluation_wall_s": 0.026139497756958008}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17273856, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s0.json b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s0.json new file mode 100644 index 0000000..b150cf5 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d1_t1_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7144275903701782}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.552, "eval_loss": 0.606060791015625, "wall_s": 4.108927488327026, "val_acc": 0.552, "val_loss": 0.606060791015625}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 4.078335285186768, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 4.078335285186768, "evaluation_wall_s": 0.02516484260559082}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17273856, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s1.json b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s1.json new file mode 100644 index 0000000..ff01ae7 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d1_t1_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6918616890907288}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5375, "eval_loss": 0.6015696411132813, "wall_s": 3.765930652618408, "val_acc": 0.5375, "val_loss": 0.6015696411132813}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 3.741680145263672, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 3.741680145263672, "evaluation_wall_s": 0.019041776657104492}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17273856, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s2.json b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s2.json new file mode 100644 index 0000000..9842dce --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d1_t1_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.70915687084198}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5215, "eval_loss": 0.6049201965332032, "wall_s": 4.112473487854004, "val_acc": 0.5215, "val_loss": 0.6049201965332032}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 4.086016654968262, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 4.086016654968262, "evaluation_wall_s": 0.02113199234008789}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17273856, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s3.json b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s3.json new file mode 100644 index 0000000..7bcb363 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d1_t1_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7197579741477966}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5485, "eval_loss": 0.6061341857910156, "wall_s": 4.424393653869629, "val_acc": 0.5485, "val_loss": 0.6061341857910156}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 4.395883560180664, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 4.395883560180664, "evaluation_wall_s": 0.0230100154876709}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17273856, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s4.json b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s4.json new file mode 100644 index 0000000..bc35b00 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t1_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d1_t1_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7674304842948914}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.523, "eval_loss": 0.5936953430175781, "wall_s": 3.994764804840088, "val_acc": 0.523, "val_loss": 0.5936953430175781}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 3.964716911315918, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 3.964716911315918, "evaluation_wall_s": 0.024221181869506836}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17273856, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s0.json b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s0.json new file mode 100644 index 0000000..08a2059 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d1_t2_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7000930905342102}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.6135, "eval_loss": 0.6196617736816407, "wall_s": 4.115383148193359, "val_acc": 0.6135, "val_loss": 0.6196617736816407}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 4.088064193725586, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 4.088064193725586, "evaluation_wall_s": 0.021833419799804688}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17273856, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s1.json b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s1.json new file mode 100644 index 0000000..6431544 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d1_t2_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6932498812675476}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.6295, "eval_loss": 0.6192972412109375, "wall_s": 4.009688377380371, "val_acc": 0.6295, "val_loss": 0.6192972412109375}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 3.979783773422241, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 3.979783773422241, "evaluation_wall_s": 0.024413585662841797}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17273856, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s2.json b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s2.json new file mode 100644 index 0000000..1dc68c9 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d1_t2_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6975919604301453}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.6295, "eval_loss": 0.6216668701171875, "wall_s": 4.2174272537231445, "val_acc": 0.6295, "val_loss": 0.6216668701171875}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 4.190459489822388, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 4.190459489822388, "evaluation_wall_s": 0.021414995193481445}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17273856, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s3.json b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s3.json new file mode 100644 index 0000000..ddf7ec2 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d1_t2_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7028841972351074}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.6225, "eval_loss": 0.6220227355957031, "wall_s": 4.13748574256897, "val_acc": 0.6225, "val_loss": 0.6220227355957031}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 4.111098289489746, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 4.111098289489746, "evaluation_wall_s": 0.021057605743408203}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17273856, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s4.json b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s4.json new file mode 100644 index 0000000..44c798a --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d1_t2_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d1_t2_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.8385413289070129}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.613, "eval_loss": 0.6234770202636719, "wall_s": 4.095953464508057, "val_acc": 0.613, "val_loss": 0.6234770202636719}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 4.069920301437378, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 4.069920301437378, "evaluation_wall_s": 0.02094435691833496}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17273856, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s0.json b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s0.json new file mode 100644 index 0000000..8b02309 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d4_t0_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6903697848320007}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.9735, "eval_loss": 0.059503128051757816, "wall_s": 7.433874130249023, "val_acc": 0.9735, "val_loss": 0.059503128051757816, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.5531695485115051, 0.6390233635902405, 0.5977039933204651], "lesion_eval_acc": 0.565, "lesion_eval_loss": 2.022367248535156, "lesion_acc_drop": 0.4085000000000001}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 7.356062173843384, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 7.356062173843384, "evaluation_wall_s": 0.07245874404907227}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17529344, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s1.json b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s1.json new file mode 100644 index 0000000..cf7b086 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d4_t0_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.749412477016449}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.9715, "eval_loss": 0.061364727020263674, "wall_s": 7.937608003616333, "val_acc": 0.9715, "val_loss": 0.061364727020263674, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.782924473285675, 1.0839496850967407, 0.6825138330459595], "lesion_eval_acc": 0.6645, "lesion_eval_loss": 0.7497943115234375, "lesion_acc_drop": 0.30700000000000005}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 7.857497215270996, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 7.857497215270996, "evaluation_wall_s": 0.07405948638916016}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17529344, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s2.json b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s2.json new file mode 100644 index 0000000..600f4b5 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d4_t0_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6906556487083435}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.9625, "eval_loss": 0.08926729583740234, "wall_s": 7.803781270980835, "val_acc": 0.9625, "val_loss": 0.08926729583740234, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.9670792818069458, 0.4067908227443695, 0.5520918369293213], "lesion_eval_acc": 0.766, "lesion_eval_loss": 2.0403499755859373, "lesion_acc_drop": 0.1965}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 7.724065542221069, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 7.724065542221069, "evaluation_wall_s": 0.07398676872253418}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17529344, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s3.json b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s3.json new file mode 100644 index 0000000..24f877e --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d4_t0_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7026023864746094}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.9615, "eval_loss": 0.11007212829589844, "wall_s": 8.405820846557617, "val_acc": 0.9615, "val_loss": 0.11007212829589844, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [1.7667630910873413, 0.2349170297384262, 1.0004180669784546], "lesion_eval_acc": 0.5125, "lesion_eval_loss": 2.780086181640625, "lesion_acc_drop": 0.44900000000000007}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 8.32558798789978, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 8.32558798789978, "evaluation_wall_s": 0.0744168758392334}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17529344, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s4.json b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s4.json new file mode 100644 index 0000000..851c58e --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t0_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d4_t0_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 1.1743755340576172}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.741, "eval_loss": 0.43585313415527344, "wall_s": 8.044363737106323, "val_acc": 0.741, "val_loss": 0.43585313415527344, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.5413612127304077, 0.1094466969370842, 0.11508037894964218], "lesion_eval_acc": 0.756, "lesion_eval_loss": 0.4506159210205078, "lesion_acc_drop": -0.015000000000000013}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 7.965662956237793, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 7.965662956237793, "evaluation_wall_s": 0.07288289070129395}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17529344, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s0.json b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s0.json new file mode 100644 index 0000000..8597ef7 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d4_t1_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.686114490032196}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.9805, "eval_loss": 0.043616046905517576, "wall_s": 7.620985984802246, "val_acc": 0.9805, "val_loss": 0.043616046905517576, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.49482274055480957, 0.5972866415977478, 0.430756539106369], "lesion_eval_acc": 0.63, "lesion_eval_loss": 0.8711539916992187, "lesion_acc_drop": 0.35050000000000003}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 7.542865753173828, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 7.542865753173828, "evaluation_wall_s": 0.0722663402557373}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17529344, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s1.json b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s1.json new file mode 100644 index 0000000..840627f --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d4_t1_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7337294220924377}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.971, "eval_loss": 0.05356445693969727, "wall_s": 8.382384777069092, "val_acc": 0.971, "val_loss": 0.05356445693969727, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.7704880237579346, 0.8600002527236938, 0.705874502658844], "lesion_eval_acc": 0.635, "lesion_eval_loss": 0.7355263977050781, "lesion_acc_drop": 0.33599999999999997}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 8.300485610961914, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 8.300485610961914, "evaluation_wall_s": 0.0754241943359375}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17529344, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s2.json b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s2.json new file mode 100644 index 0000000..a72604c --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d4_t1_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6891603469848633}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.9625, "eval_loss": 0.08642630386352539, "wall_s": 7.85395359992981, "val_acc": 0.9625, "val_loss": 0.08642630386352539, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.9557654857635498, 0.44688287377357483, 0.8256698846817017], "lesion_eval_acc": 0.5955, "lesion_eval_loss": 3.7492200927734376, "lesion_acc_drop": 0.367}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 7.775715112686157, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 7.775715112686157, "evaluation_wall_s": 0.07197785377502441}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17529344, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s3.json b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s3.json new file mode 100644 index 0000000..bd0dd41 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d4_t1_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7055310606956482}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.9585, "eval_loss": 0.10342636108398437, "wall_s": 7.842729568481445, "val_acc": 0.9585, "val_loss": 0.10342636108398437, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [2.017836809158325, 0.5820943713188171, 0.49246856570243835], "lesion_eval_acc": 0.647, "lesion_eval_loss": 3.02464404296875, "lesion_acc_drop": 0.3115}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 7.759349346160889, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 7.759349346160889, "evaluation_wall_s": 0.0773928165435791}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17529344, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s4.json b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s4.json new file mode 100644 index 0000000..6fdad80 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t1_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d4_t1_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 1.2137476205825806}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.691, "eval_loss": 0.5254037170410156, "wall_s": 8.235943794250488, "val_acc": 0.691, "val_loss": 0.5254037170410156, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.6370808482170105, 0.05066376551985741, 0.11072749644517899], "lesion_eval_acc": 0.5505, "lesion_eval_loss": 0.5588606567382812, "lesion_acc_drop": 0.14049999999999996}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 8.164164304733276, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 8.164164304733276, "evaluation_wall_s": 0.06555461883544922}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17529344, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s0.json b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s0.json new file mode 100644 index 0000000..9b7e7d4 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d4_t2_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7004839181900024}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.9525, "eval_loss": 0.12547859954833984, "wall_s": 7.946804046630859, "val_acc": 0.9525, "val_loss": 0.12547859954833984, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.548296332359314, 0.6270938515663147, 0.4090343713760376], "lesion_eval_acc": 0.604, "lesion_eval_loss": 1.8691396484375, "lesion_acc_drop": 0.34850000000000003}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 7.8710036277771, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 7.8710036277771, "evaluation_wall_s": 0.07027864456176758}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17529344, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s1.json b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s1.json new file mode 100644 index 0000000..f87385d --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d4_t2_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7904043793678284}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.9425, "eval_loss": 0.18557228088378908, "wall_s": 8.315591096878052, "val_acc": 0.9425, "val_loss": 0.18557228088378908, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.6970807909965515, 0.9083057641983032, 0.7748426198959351], "lesion_eval_acc": 0.6925, "lesion_eval_loss": 1.3189234008789064, "lesion_acc_drop": 0.25}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 8.236629009246826, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 8.236629009246826, "evaluation_wall_s": 0.07266068458557129}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17529344, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s2.json b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s2.json new file mode 100644 index 0000000..283a565 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d4_t2_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6941237449645996}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.938, "eval_loss": 0.18500452423095703, "wall_s": 8.13248610496521, "val_acc": 0.938, "val_loss": 0.18500452423095703, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.96188884973526, 0.9660025835037231, 0.49828121066093445], "lesion_eval_acc": 0.6315, "lesion_eval_loss": 3.3104510498046875, "lesion_acc_drop": 0.3065}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 8.054811716079712, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 8.054811716079712, "evaluation_wall_s": 0.07181501388549805}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17529344, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s3.json b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s3.json new file mode 100644 index 0000000..f2fb156 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d4_t2_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.696068286895752}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.9775, "eval_loss": 0.0666396484375, "wall_s": 7.810996055603027, "val_acc": 0.9775, "val_loss": 0.0666396484375, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [1.788214921951294, 0.581813395023346, 0.37464359402656555], "lesion_eval_acc": 0.476, "lesion_eval_loss": 2.7538897705078127, "lesion_acc_drop": 0.5015000000000001}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 7.733540773391724, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 7.733540773391724, "evaluation_wall_s": 0.07139801979064941}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17529344, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s4.json b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s4.json new file mode 100644 index 0000000..06589a7 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_bp_d4_t2_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "bp", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_bp_d4_t2_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 1.1055001020431519}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.884, "eval_loss": 0.3387477569580078, "wall_s": 8.702982425689697, "val_acc": 0.884, "val_loss": 0.3387477569580078, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.9640625715255737, 0.23909826576709747, 0.3484726548194885], "lesion_eval_acc": 0.6185, "lesion_eval_loss": 3.1072716064453125, "lesion_acc_drop": 0.26549999999999996}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 8.624804735183716, "diagnostics_wall_s": 0.0, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 8.624804735183716, "evaluation_wall_s": 0.07232046127319336}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17529344, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s0.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s0.json new file mode 100644 index 0000000..43bf755 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7099499106407166}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.63, "eval_loss": 0.6268773498535156, "wall_s": 3.2204158306121826, "val_acc": 0.63, "val_loss": 0.6268773498535156, "cos_r_negg": [0.6381222009658813], "cos_innovation_negg": [0.6381222009658813], "cos_apical_negg": [0.6381222009658813], "cos_Ac_negg": [0.6381222009658813], "r_norm": [0.7443219423294067], "g_norm": [0.006909060291945934], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 3.1170079708099365, "diagnostics_wall_s": 0.07257676124572754, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 3.1170079708099365, "evaluation_wall_s": 0.025905370712280273}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s1.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s1.json new file mode 100644 index 0000000..e7e71f9 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6922826766967773}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5, "eval_loss": 0.69442529296875, "wall_s": 2.755969285964966, "val_acc": 0.5, "val_loss": 0.69442529296875, "cos_r_negg": [0.0422159880399704], "cos_innovation_negg": [0.0422159880399704], "cos_apical_negg": [0.0422159880399704], "cos_Ac_negg": [0.0422159880399704], "r_norm": [1.6446845531463623], "g_norm": [0.0014385026879608631], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.6598446369171143, "diagnostics_wall_s": 0.06754660606384277, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.6598446369171143, "evaluation_wall_s": 0.02388167381286621}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s2.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s2.json new file mode 100644 index 0000000..33114e4 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7055632472038269}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5685, "eval_loss": 0.6125427551269531, "wall_s": 2.6697611808776855, "val_acc": 0.5685, "val_loss": 0.6125427551269531, "cos_r_negg": [0.7954922318458557], "cos_innovation_negg": [0.7954922318458557], "cos_apical_negg": [0.7954922318458557], "cos_Ac_negg": [0.7954922318458557], "r_norm": [1.0125494003295898], "g_norm": [0.006798403803259134], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.566847562789917, "diagnostics_wall_s": 0.06678414344787598, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.566847562789917, "evaluation_wall_s": 0.0315394401550293}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s3.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s3.json new file mode 100644 index 0000000..ab822ed --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7144617438316345}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5775, "eval_loss": 0.6774878540039062, "wall_s": 2.7053627967834473, "val_acc": 0.5775, "val_loss": 0.6774878540039062, "cos_r_negg": [0.4672079086303711], "cos_innovation_negg": [0.4672079086303711], "cos_apical_negg": [0.4672079086303711], "cos_Ac_negg": [0.4672079086303711], "r_norm": [0.6758668422698975], "g_norm": [0.0032122500706464052], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.6050524711608887, "diagnostics_wall_s": 0.07041215896606445, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.6050524711608887, "evaluation_wall_s": 0.024814367294311523}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s4.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s4.json new file mode 100644 index 0000000..ebe5d26 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d1_t0_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7867429256439209}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.7295, "eval_loss": 0.5244359130859375, "wall_s": 2.6169021129608154, "val_acc": 0.7295, "val_loss": 0.5244359130859375, "cos_r_negg": [0.8994601368904114], "cos_innovation_negg": [0.8994601368904114], "cos_apical_negg": [0.8994601368904114], "cos_Ac_negg": [0.8994601368904114], "r_norm": [1.4971390962600708], "g_norm": [0.009177722968161106], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.505246162414551, "diagnostics_wall_s": 0.08174443244934082, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.505246162414551, "evaluation_wall_s": 0.025374412536621094}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s0.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s0.json new file mode 100644 index 0000000..4484077 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7144274711608887}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.3775, "eval_loss": 0.6937947998046875, "wall_s": 2.69585919380188, "val_acc": 0.3775, "val_loss": 0.6937947998046875, "cos_r_negg": [-0.6537743806838989], "cos_innovation_negg": [-0.6537743806838989], "cos_apical_negg": [-0.6537743806838989], "cos_Ac_negg": [-0.6537743806838989], "r_norm": [0.807400107383728], "g_norm": [0.001917891320772469], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.5975186824798584, "diagnostics_wall_s": 0.06919360160827637, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.5975186824798584, "evaluation_wall_s": 0.023614168167114258}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s1.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s1.json new file mode 100644 index 0000000..1914653 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.691861629486084}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5285, "eval_loss": 0.6940193176269531, "wall_s": 2.9417903423309326, "val_acc": 0.5285, "val_loss": 0.6940193176269531, "cos_r_negg": [0.06625309586524963], "cos_innovation_negg": [0.06625309586524963], "cos_apical_negg": [0.06625309586524963], "cos_Ac_negg": [0.06625309586524963], "r_norm": [1.6463780403137207], "g_norm": [0.0014416410122066736], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.835766077041626, "diagnostics_wall_s": 0.06854748725891113, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.835766077041626, "evaluation_wall_s": 0.03248786926269531}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s2.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s2.json new file mode 100644 index 0000000..0c8f8c9 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7091569304466248}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.437, "eval_loss": 0.6936712646484375, "wall_s": 2.550065755844116, "val_acc": 0.437, "val_loss": 0.6936712646484375, "cos_r_negg": [-0.8953896760940552], "cos_innovation_negg": [-0.8953896760940552], "cos_apical_negg": [-0.8953896760940552], "cos_Ac_negg": [-0.8953896760940552], "r_norm": [1.1212096214294434], "g_norm": [0.0014783325605094433], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.457213878631592, "diagnostics_wall_s": 0.06447005271911621, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.457213878631592, "evaluation_wall_s": 0.023493051528930664}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s3.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s3.json new file mode 100644 index 0000000..915fdc3 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7197580337524414}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.4005, "eval_loss": 0.6886610107421876, "wall_s": 2.669794797897339, "val_acc": 0.4005, "val_loss": 0.6886610107421876, "cos_r_negg": [0.49667543172836304], "cos_innovation_negg": [0.49667543172836304], "cos_apical_negg": [0.49667543172836304], "cos_Ac_negg": [0.49667543172836304], "r_norm": [0.6817830801010132], "g_norm": [0.0013702992582693696], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.564960479736328, "diagnostics_wall_s": 0.0760040283203125, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.564960479736328, "evaluation_wall_s": 0.024204730987548828}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s4.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s4.json new file mode 100644 index 0000000..d279bf6 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d1_t1_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7674305438995361}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.721, "eval_loss": 0.5305926208496093, "wall_s": 2.8444759845733643, "val_acc": 0.721, "val_loss": 0.5305926208496093, "cos_r_negg": [0.9022473096847534], "cos_innovation_negg": [0.9022473096847534], "cos_apical_negg": [0.9022473096847534], "cos_Ac_negg": [0.9022473096847534], "r_norm": [1.494638204574585], "g_norm": [0.009266329929232597], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.7179019451141357, "diagnostics_wall_s": 0.0899193286895752, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.7179019451141357, "evaluation_wall_s": 0.031585693359375}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s0.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s0.json new file mode 100644 index 0000000..37ca8aa --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.700093150138855}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5125, "eval_loss": 0.6401134033203125, "wall_s": 2.5725858211517334, "val_acc": 0.5125, "val_loss": 0.6401134033203125, "cos_r_negg": [0.6190553307533264], "cos_innovation_negg": [0.6190553307533264], "cos_apical_negg": [0.6190553307533264], "cos_Ac_negg": [0.6190553307533264], "r_norm": [0.755313515663147], "g_norm": [0.006378652527928352], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.472454071044922, "diagnostics_wall_s": 0.07224225997924805, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.472454071044922, "evaluation_wall_s": 0.023609638214111328}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s1.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s1.json new file mode 100644 index 0000000..8704409 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6932498216629028}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5, "eval_loss": 0.6937125244140625, "wall_s": 2.8784992694854736, "val_acc": 0.5, "val_loss": 0.6937125244140625, "cos_r_negg": [0.0021840184926986694], "cos_innovation_negg": [0.0021840184926986694], "cos_apical_negg": [0.0021840184926986694], "cos_Ac_negg": [0.0021840184926986694], "r_norm": [1.6410367488861084], "g_norm": [0.0014340589987114072], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.7704505920410156, "diagnostics_wall_s": 0.0716547966003418, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.7704505920410156, "evaluation_wall_s": 0.03153657913208008}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s2.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s2.json new file mode 100644 index 0000000..5e06983 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6975919008255005}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5105, "eval_loss": 0.6253689575195313, "wall_s": 2.754265546798706, "val_acc": 0.5105, "val_loss": 0.6253689575195313, "cos_r_negg": [0.7872234582901001], "cos_innovation_negg": [0.7872234582901001], "cos_apical_negg": [0.7872234582901001], "cos_Ac_negg": [0.7872234582901001], "r_norm": [1.022896409034729], "g_norm": [0.006786737125366926], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.6610193252563477, "diagnostics_wall_s": 0.06567668914794922, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.6610193252563477, "evaluation_wall_s": 0.023148298263549805}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s3.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s3.json new file mode 100644 index 0000000..4619649 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7028841376304626}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.617, "eval_loss": 0.6814989929199219, "wall_s": 2.741743326187134, "val_acc": 0.617, "val_loss": 0.6814989929199219, "cos_r_negg": [0.41905343532562256], "cos_innovation_negg": [0.41905343532562256], "cos_apical_negg": [0.41905343532562256], "cos_Ac_negg": [0.41905343532562256], "r_norm": [0.6754723787307739], "g_norm": [0.003080698661506176], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.64243745803833, "diagnostics_wall_s": 0.06916308403015137, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.64243745803833, "evaluation_wall_s": 0.025385141372680664}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s4.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s4.json new file mode 100644 index 0000000..916509b --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d1_t2_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.8385413885116577}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.7175, "eval_loss": 0.5380291137695312, "wall_s": 2.831052541732788, "val_acc": 0.7175, "val_loss": 0.5380291137695312, "cos_r_negg": [0.8999611735343933], "cos_innovation_negg": [0.8999611735343933], "cos_apical_negg": [0.8999611735343933], "cos_Ac_negg": [0.8999611735343933], "r_norm": [1.5154743194580078], "g_norm": [0.009075421839952469], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.7259016036987305, "diagnostics_wall_s": 0.07642984390258789, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.7259016036987305, "evaluation_wall_s": 0.023874282836914062}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s0.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s0.json new file mode 100644 index 0000000..98e5859 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6903698444366455}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.893, "eval_loss": 0.25367149353027346, "wall_s": 5.632862329483032, "val_acc": 0.893, "val_loss": 0.25367149353027346, "cos_r_negg": [0.9387437105178833, 0.8077287673950195, 0.867480993270874, 0.4752769470214844], "cos_innovation_negg": [0.9387437105178833, 0.8077287673950195, 0.867480993270874, 0.4752769470214844], "cos_apical_negg": [0.9387437105178833, 0.8077287673950195, 0.867480993270874, 0.4752769470214844], "cos_Ac_negg": [0.9387437105178833, 0.8077287673950195, 0.867480993270874, 0.4752769470214844], "r_norm": [0.6103125810623169, 0.4610728323459625, 0.39773792028427124, 0.5367587208747864], "g_norm": [0.01377161219716072, 0.01218781340867281, 0.008680791594088078, 0.006580328568816185], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.4212634861469269, 0.9535810351371765, 0.5171565413475037], "lesion_eval_acc": 0.5, "lesion_eval_loss": 3.4355047607421874, "lesion_acc_drop": 0.393}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.503223419189453, "diagnostics_wall_s": 0.08867502212524414, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.503223419189453, "evaluation_wall_s": 0.03630495071411133}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s1.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s1.json new file mode 100644 index 0000000..0a1ebd9 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7494125366210938}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.548, "eval_loss": 0.6879402160644531, "wall_s": 5.701672315597534, "val_acc": 0.548, "val_loss": 0.6879402160644531, "cos_r_negg": [0.32946598529815674, -0.22963553667068481, -0.9623533487319946, 0.5182471871376038], "cos_innovation_negg": [0.32946598529815674, -0.22963553667068481, -0.9623533487319946, 0.5182471871376038], "cos_apical_negg": [0.32946598529815674, -0.22963553667068481, -0.9623533487319946, 0.5182471871376038], "cos_Ac_negg": [0.32946598529815674, -0.22963553667068481, -0.9623533487319946, 0.5182471871376038], "r_norm": [1.9099602699279785, 1.1090948581695557, 1.9587345123291016, 1.1576108932495117], "g_norm": [0.002804109826683998, 0.0027241313364356756, 0.0026775142177939415, 0.0028336055111140013], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.19062604010105133, 0.564626932144165, 0.8382281064987183], "lesion_eval_acc": 0.442, "lesion_eval_loss": 0.85650634765625, "lesion_acc_drop": 0.10600000000000004}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.567786455154419, "diagnostics_wall_s": 0.09290814399719238, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.567786455154419, "evaluation_wall_s": 0.03613543510437012}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s2.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s2.json new file mode 100644 index 0000000..69002db --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6906555891036987}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.6045, "eval_loss": 0.5727956848144531, "wall_s": 5.940492630004883, "val_acc": 0.6045, "val_loss": 0.5727956848144531, "cos_r_negg": [0.9201407432556152, -0.5260127782821655, 0.22619004547595978, 0.8064072132110596], "cos_innovation_negg": [0.9201407432556152, -0.5260127782821655, 0.22619004547595978, 0.8064072132110596], "cos_apical_negg": [0.9201407432556152, -0.5260127782821655, 0.22619004547595978, 0.8064072132110596], "cos_Ac_negg": [0.9201407432556152, -0.5260127782821655, 0.22619004547595978, 0.8064072132110596], "r_norm": [0.8368968963623047, 1.428114891052246, 0.942912220954895, 1.008380651473999], "g_norm": [0.012037801556289196, 0.014238949865102768, 0.011532217264175415, 0.007170531898736954], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.5954912900924683, 0.6569982171058655, 1.1329480409622192], "lesion_eval_acc": 0.4085, "lesion_eval_loss": 0.8772562866210938, "lesion_acc_drop": 0.19600000000000006}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.800463914871216, "diagnostics_wall_s": 0.09128975868225098, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.800463914871216, "evaluation_wall_s": 0.043892621994018555}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s3.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s3.json new file mode 100644 index 0000000..b9fe091 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7026023268699646}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.94, "eval_loss": 0.25042525482177735, "wall_s": 5.571286916732788, "val_acc": 0.94, "val_loss": 0.25042525482177735, "cos_r_negg": [0.1879110187292099, 0.4076645076274872, 0.9121440649032593, 0.9679760932922363], "cos_innovation_negg": [0.1879110187292099, 0.4076645076274872, 0.9121440649032593, 0.9679760932922363], "cos_apical_negg": [0.1879110187292099, 0.4076645076274872, 0.9121440649032593, 0.9679760932922363], "cos_Ac_negg": [0.1879110187292099, 0.4076645076274872, 0.9121440649032593, 0.9679760932922363], "r_norm": [0.5813260078430176, 0.5943952202796936, 0.813428521156311, 0.43674609065055847], "g_norm": [0.021987497806549072, 0.021553903818130493, 0.015678197145462036, 0.00656924257054925], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.6310636401176453, 0.22090193629264832, 0.2915653884410858], "lesion_eval_acc": 0.6175, "lesion_eval_loss": 0.6150129699707031, "lesion_acc_drop": 0.3224999999999999}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.442969560623169, "diagnostics_wall_s": 0.08774399757385254, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.442969560623169, "evaluation_wall_s": 0.03595447540283203}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s4.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s4.json new file mode 100644 index 0000000..4dae84a --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d4_t0_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 1.1743755340576172}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.6645, "eval_loss": 0.4307123565673828, "wall_s": 5.590679168701172, "val_acc": 0.6645, "val_loss": 0.4307123565673828, "cos_r_negg": [0.35997045040130615, -0.09435587376356125, -0.32991841435432434, 0.8966705799102783], "cos_innovation_negg": [0.35997045040130615, -0.09435587376356125, -0.32991841435432434, 0.8966705799102783], "cos_apical_negg": [0.35997045040130615, -0.09435587376356125, -0.32991841435432434, 0.8966705799102783], "cos_Ac_negg": [0.35997045040130615, -0.09435587376356125, -0.32991841435432434, 0.8966705799102783], "r_norm": [0.8576339483261108, 0.901697039604187, 1.0227551460266113, 1.42327880859375], "g_norm": [0.019810788333415985, 0.016717838123440742, 0.013885117135941982, 0.007210355252027512], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.4204072952270508, 0.0797010287642479, 0.17843063175678253], "lesion_eval_acc": 0.5225, "lesion_eval_loss": 1.140498779296875, "lesion_acc_drop": 0.14200000000000002}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.44824481010437, "diagnostics_wall_s": 0.09948515892028809, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.44824481010437, "evaluation_wall_s": 0.0365903377532959}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s0.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s0.json new file mode 100644 index 0000000..2bda202 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.686114490032196}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.933, "eval_loss": 0.23948320770263673, "wall_s": 5.507511377334595, "val_acc": 0.933, "val_loss": 0.23948320770263673, "cos_r_negg": [0.9356802701950073, 0.7872922420501709, 0.8644906282424927, 0.486294686794281], "cos_innovation_negg": [0.9356802701950073, 0.7872922420501709, 0.8644906282424927, 0.486294686794281], "cos_apical_negg": [0.9356802701950073, 0.7872922420501709, 0.8644906282424927, 0.486294686794281], "cos_Ac_negg": [0.9356802701950073, 0.7872922420501709, 0.8644906282424927, 0.486294686794281], "r_norm": [0.5507198572158813, 0.41605231165885925, 0.35890159010887146, 0.484347939491272], "g_norm": [0.011479105800390244, 0.010674834251403809, 0.007763249799609184, 0.005922361277043819], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.4174897372722626, 0.9466758966445923, 0.5357286930084229], "lesion_eval_acc": 0.5, "lesion_eval_loss": 3.1990943603515625, "lesion_acc_drop": 0.43300000000000005}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.369952440261841, "diagnostics_wall_s": 0.09671878814697266, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.369952440261841, "evaluation_wall_s": 0.036057472229003906}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s1.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s1.json new file mode 100644 index 0000000..b82afe9 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7337294816970825}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.731, "eval_loss": 0.603870849609375, "wall_s": 5.889064788818359, "val_acc": 0.731, "val_loss": 0.603870849609375, "cos_r_negg": [0.07503455132246017, 0.8850080370903015, 0.8480286002159119, 0.49959999322891235], "cos_innovation_negg": [0.07503455132246017, 0.8850080370903015, 0.8480286002159119, 0.49959999322891235], "cos_apical_negg": [0.07503455132246017, 0.8850080370903015, 0.8480286002159119, 0.49959999322891235], "cos_Ac_negg": [0.07503455132246017, 0.8850080370903015, 0.8480286002159119, 0.49959999322891235], "r_norm": [1.6789445877075195, 0.9749464988708496, 1.7218196392059326, 1.017594337463379], "g_norm": [0.015472279861569405, 0.018419174477458, 0.010109108872711658, 0.006710904184728861], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.32275137305259705, 0.5108727216720581, 0.6919687986373901], "lesion_eval_acc": 0.5, "lesion_eval_loss": 1.2617421875, "lesion_acc_drop": 0.23099999999999998}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.7617621421813965, "diagnostics_wall_s": 0.08734726905822754, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.7617621421813965, "evaluation_wall_s": 0.03534722328186035}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s2.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s2.json new file mode 100644 index 0000000..275c27a --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6891602873802185}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.728, "eval_loss": 0.5929824523925781, "wall_s": 5.870525598526001, "val_acc": 0.728, "val_loss": 0.5929824523925781, "cos_r_negg": [0.9243294596672058, -0.5396212339401245, 0.22078830003738403, 0.8045492768287659], "cos_innovation_negg": [0.9243294596672058, -0.5396212339401245, 0.22078830003738403, 0.8045492768287659], "cos_apical_negg": [0.9243294596672058, -0.5396212339401245, 0.22078830003738403, 0.8045492768287659], "cos_Ac_negg": [0.9243294596672058, -0.5396212339401245, 0.22078830003738403, 0.8045492768287659], "r_norm": [0.8272742033004761, 1.4116944074630737, 0.9320707321166992, 0.9967862963676453], "g_norm": [0.011796096339821815, 0.014009530656039715, 0.01140005886554718, 0.007074049673974514], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.5936934351921082, 0.6572055220603943, 1.141676664352417], "lesion_eval_acc": 0.4315, "lesion_eval_loss": 0.8801928100585937, "lesion_acc_drop": 0.2965}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.743980884552002, "diagnostics_wall_s": 0.08605074882507324, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.743980884552002, "evaluation_wall_s": 0.035198211669921875}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s3.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s3.json new file mode 100644 index 0000000..e2c49c0 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.705531120300293}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.778, "eval_loss": 0.4777294921875, "wall_s": 8.051000356674194, "val_acc": 0.778, "val_loss": 0.4777294921875, "cos_r_negg": [0.30178302526474, 0.3580387830734253, 0.9626747965812683, 0.8955805897712708], "cos_innovation_negg": [0.30178302526474, 0.3580387830734253, 0.9626747965812683, 0.8955805897712708], "cos_apical_negg": [0.30178302526474, 0.3580387830734253, 0.9626747965812683, 0.8955805897712708], "cos_Ac_negg": [0.30178302526474, 0.3580387830734253, 0.9626747965812683, 0.8955805897712708], "r_norm": [0.9930284023284912, 1.0153534412384033, 1.3895089626312256, 0.7460551857948303], "g_norm": [0.00999067910015583, 0.01050235703587532, 0.009072241373360157, 0.005770443007349968], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.7697548270225525, 0.23899975419044495, 0.4224521219730377], "lesion_eval_acc": 0.539, "lesion_eval_loss": 0.60719677734375, "lesion_acc_drop": 0.239}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 7.908400774002075, "diagnostics_wall_s": 0.09749126434326172, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 7.908400774002075, "evaluation_wall_s": 0.03838086128234863}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s4.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s4.json new file mode 100644 index 0000000..19d1252 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d4_t1_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 1.213747501373291}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.605, "eval_loss": 0.6568758850097656, "wall_s": 6.0349812507629395, "val_acc": 0.605, "val_loss": 0.6568758850097656, "cos_r_negg": [0.4690897464752197, -0.13500896096229553, -0.08921871334314346, 0.8651923537254333], "cos_innovation_negg": [0.4690897464752197, -0.13500896096229553, -0.08921871334314346, 0.8651923537254333], "cos_apical_negg": [0.4690897464752197, -0.13500896096229553, -0.08921871334314346, 0.8651923537254333], "cos_Ac_negg": [0.4690897464752197, -0.13500896096229553, -0.08921871334314346, 0.8651923537254333], "r_norm": [1.3212354183197021, 1.3891171216964722, 1.5756144523620605, 2.1926445960998535], "g_norm": [0.002936787437647581, 0.002642173320055008, 0.0023986161686480045, 0.00221154373139143], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.45367515087127686, 0.2049386203289032, 0.12880940735340118], "lesion_eval_acc": 0.6525, "lesion_eval_loss": 0.6695528259277344, "lesion_acc_drop": -0.04749999999999999}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.895235061645508, "diagnostics_wall_s": 0.09023880958557129, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.895235061645508, "evaluation_wall_s": 0.04458951950073242}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s0.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s0.json new file mode 100644 index 0000000..f7a6020 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7004837989807129}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.908, "eval_loss": 0.2594555740356445, "wall_s": 5.7472875118255615, "val_acc": 0.908, "val_loss": 0.2594555740356445, "cos_r_negg": [0.9543294906616211, 0.824755072593689, 0.41177552938461304, 0.35651978850364685], "cos_innovation_negg": [0.9543294906616211, 0.824755072593689, 0.41177552938461304, 0.35651978850364685], "cos_apical_negg": [0.9543294906616211, 0.824755072593689, 0.41177552938461304, 0.35651978850364685], "cos_Ac_negg": [0.9543294906616211, 0.824755072593689, 0.41177552938461304, 0.35651978850364685], "r_norm": [0.6473711729049683, 0.48906949162483215, 0.4218888282775879, 0.569351077079773], "g_norm": [0.0170056764036417, 0.014962108805775642, 0.008503971621394157, 0.006803838070482016], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.4491223096847534, 0.9143012762069702, 0.5318959951400757], "lesion_eval_acc": 0.5, "lesion_eval_loss": 2.6199085693359376, "lesion_acc_drop": 0.40800000000000003}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.612735271453857, "diagnostics_wall_s": 0.08664608001708984, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.612735271453857, "evaluation_wall_s": 0.04319906234741211}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s1.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s1.json new file mode 100644 index 0000000..939a63d --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7904044985771179}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.4935, "eval_loss": 0.6343562316894531, "wall_s": 5.867340564727783, "val_acc": 0.4935, "val_loss": 0.6343562316894531, "cos_r_negg": [0.28000959753990173, -0.5391826629638672, -0.5774836540222168, 0.6482595205307007], "cos_innovation_negg": [0.28000959753990173, -0.5391826629638672, -0.5774836540222168, 0.6482595205307007], "cos_apical_negg": [0.28000959753990173, -0.5391826629638672, -0.5774836540222168, 0.6482595205307007], "cos_Ac_negg": [0.28000959753990173, -0.5391826629638672, -0.5774836540222168, 0.6482595205307007], "r_norm": [1.7823982238769531, 1.0350209474563599, 1.8279149532318115, 1.080296516418457], "g_norm": [0.007991647347807884, 0.007442602887749672, 0.007383465301245451, 0.007306929212063551], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.19296739995479584, 0.5819979906082153, 0.8941615223884583], "lesion_eval_acc": 0.447, "lesion_eval_loss": 0.9224317626953125, "lesion_acc_drop": 0.046499999999999986}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.722196102142334, "diagnostics_wall_s": 0.09688520431518555, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.722196102142334, "evaluation_wall_s": 0.0435025691986084}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s2.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s2.json new file mode 100644 index 0000000..6c94c5c --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6941238045692444}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.696, "eval_loss": 0.5946675720214843, "wall_s": 5.38844895362854, "val_acc": 0.696, "val_loss": 0.5946675720214843, "cos_r_negg": [0.9214762449264526, -0.508080244064331, 0.2347370982170105, 0.8050945997238159], "cos_innovation_negg": [0.9214762449264526, -0.508080244064331, 0.2347370982170105, 0.8050945997238159], "cos_apical_negg": [0.9214762449264526, -0.508080244064331, 0.2347370982170105, 0.8050945997238159], "cos_Ac_negg": [0.9214762449264526, -0.508080244064331, 0.2347370982170105, 0.8050945997238159], "r_norm": [0.8555922508239746, 1.460017442703247, 0.9639759063720703, 1.0309066772460938], "g_norm": [0.011961137875914574, 0.0142056317999959, 0.011554541997611523, 0.007152342237532139], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.5966120362281799, 0.6489737033843994, 1.1178785562515259], "lesion_eval_acc": 0.4285, "lesion_eval_loss": 0.8597620849609375, "lesion_acc_drop": 0.26749999999999996}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.25786018371582, "diagnostics_wall_s": 0.0910334587097168, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.25786018371582, "evaluation_wall_s": 0.03497481346130371}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s3.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s3.json new file mode 100644 index 0000000..45944fe --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6960683465003967}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.695, "eval_loss": 0.5937296142578125, "wall_s": 5.726406812667847, "val_acc": 0.695, "val_loss": 0.5937296142578125, "cos_r_negg": [0.34784483909606934, 0.4112607538700104, 0.3723892867565155, 0.8468643426895142], "cos_innovation_negg": [0.34784483909606934, 0.4112607538700104, 0.3723892867565155, 0.8468643426895142], "cos_apical_negg": [0.34784483909606934, 0.4112607538700104, 0.3723892867565155, 0.8468643426895142], "cos_Ac_negg": [0.34784483909606934, 0.4112607538700104, 0.3723892867565155, 0.8468643426895142], "r_norm": [1.1493905782699585, 1.1752309799194336, 1.6083009243011475, 0.8635289669036865], "g_norm": [0.014407929964363575, 0.013165452517569065, 0.010788684710860252, 0.006603921763598919], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.6147326231002808, 0.5923569798469543, 0.29215505719184875], "lesion_eval_acc": 0.5, "lesion_eval_loss": 0.726744384765625, "lesion_acc_drop": 0.19499999999999995}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.591581583023071, "diagnostics_wall_s": 0.09367799758911133, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.591581583023071, "evaluation_wall_s": 0.036510467529296875}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s4.json b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s4.json new file mode 100644 index 0000000..8aca534 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "dfa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_dfa_d4_t2_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 1.1055001020431519}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.7615, "eval_loss": 0.47405506896972655, "wall_s": 5.558012247085571, "val_acc": 0.7615, "val_loss": 0.47405506896972655, "cos_r_negg": [0.3707292675971985, 0.5508004426956177, 0.7450497150421143, 0.9531793594360352], "cos_innovation_negg": [0.3707292675971985, 0.5508004426956177, 0.7450497150421143, 0.9531793594360352], "cos_apical_negg": [0.3707292675971985, 0.5508004426956177, 0.7450497150421143, 0.9531793594360352], "cos_Ac_negg": [0.3707292675971985, 0.5508004426956177, 0.7450497150421143, 0.9531793594360352], "r_norm": [0.9581700563430786, 1.0073983669281006, 1.1426477432250977, 1.5901228189468384], "g_norm": [0.012580916285514832, 0.012274783104658127, 0.011739682406187057, 0.008216301910579205], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.47974613308906555, 0.24105918407440186, 0.2394786924123764], "lesion_eval_acc": 0.5925, "lesion_eval_loss": 1.346644287109375, "lesion_acc_drop": 0.16899999999999993}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.4291040897369385, "diagnostics_wall_s": 0.08825540542602539, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.4291040897369385, "evaluation_wall_s": 0.03618311882019043}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s0.json b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s0.json new file mode 100644 index 0000000..bb55cb4 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d1_t0_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7099499106407166}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.6225, "eval_loss": 0.6154347534179687, "wall_s": 2.6812353134155273, "val_acc": 0.6225, "val_loss": 0.6154347534179687, "cos_fa_negg": [0.6839618682861328], "loss": 0.6233125925064087}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.605117082595825, "diagnostics_wall_s": 0.04765605926513672, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.605117082595825, "evaluation_wall_s": 0.023923635482788086}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17310208, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s1.json b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s1.json new file mode 100644 index 0000000..8e475a9 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d1_t0_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6922825574874878}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.6735, "eval_loss": 0.630708984375, "wall_s": 2.5900495052337646, "val_acc": 0.6735, "val_loss": 0.630708984375, "cos_fa_negg": [0.797402024269104], "loss": 0.6372454762458801}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.5126893520355225, "diagnostics_wall_s": 0.047101497650146484, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.5126893520355225, "evaluation_wall_s": 0.025781869888305664}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17310208, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s2.json b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s2.json new file mode 100644 index 0000000..35fb909 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d1_t0_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7055633664131165}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.603, "eval_loss": 0.6184385986328125, "wall_s": 2.8671061992645264, "val_acc": 0.603, "val_loss": 0.6184385986328125, "cos_fa_negg": [0.9326964616775513], "loss": 0.6263642907142639}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.7910256385803223, "diagnostics_wall_s": 0.04670405387878418, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.7910256385803223, "evaluation_wall_s": 0.02456212043762207}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17310208, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s3.json b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s3.json new file mode 100644 index 0000000..d064894 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d1_t0_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7144617438316345}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.628, "eval_loss": 0.6275732727050781, "wall_s": 2.863197088241577, "val_acc": 0.628, "val_loss": 0.6275732727050781, "cos_fa_negg": [0.877512514591217], "loss": 0.6344802379608154}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.7694497108459473, "diagnostics_wall_s": 0.057338714599609375, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.7694497108459473, "evaluation_wall_s": 0.030971765518188477}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17310208, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s4.json b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s4.json new file mode 100644 index 0000000..f9cf957 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t0_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d1_t0_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7867427468299866}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.614, "eval_loss": 0.6171889343261718, "wall_s": 3.6563429832458496, "val_acc": 0.614, "val_loss": 0.6171889343261718, "cos_fa_negg": [0.8810242414474487], "loss": 0.6251710653305054}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 3.5815317630767822, "diagnostics_wall_s": 0.0449376106262207, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 3.5815317630767822, "evaluation_wall_s": 0.024627208709716797}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17310208, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s0.json b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s0.json new file mode 100644 index 0000000..f5f2b20 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d1_t1_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7144275903701782}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.602, "eval_loss": 0.6271330871582032, "wall_s": 2.6368539333343506, "val_acc": 0.602, "val_loss": 0.6271330871582032, "cos_fa_negg": [0.6758211255073547], "loss": 0.6241094470024109}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.556360960006714, "diagnostics_wall_s": 0.04848504066467285, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.556360960006714, "evaluation_wall_s": 0.027415990829467773}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17310208, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s1.json b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s1.json new file mode 100644 index 0000000..1ff9a43 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d1_t1_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6918616890907288}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5935, "eval_loss": 0.6352244262695312, "wall_s": 2.621917963027954, "val_acc": 0.5935, "val_loss": 0.6352244262695312, "cos_fa_negg": [0.7770487666130066], "loss": 0.632664144039154}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.5485594272613525, "diagnostics_wall_s": 0.045291900634765625, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.5485594272613525, "evaluation_wall_s": 0.02371501922607422}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17310208, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s2.json b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s2.json new file mode 100644 index 0000000..ee26e52 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d1_t1_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.70915687084198}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.6085, "eval_loss": 0.6350870361328125, "wall_s": 2.7574410438537598, "val_acc": 0.6085, "val_loss": 0.6350870361328125, "cos_fa_negg": [0.9343919157981873], "loss": 0.6327211260795593}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.6740288734436035, "diagnostics_wall_s": 0.054131269454956055, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.6740288734436035, "evaluation_wall_s": 0.02457737922668457}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17310208, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s3.json b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s3.json new file mode 100644 index 0000000..6b9aaca --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d1_t1_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7197579741477966}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.606, "eval_loss": 0.6369519348144531, "wall_s": 2.7029292583465576, "val_acc": 0.606, "val_loss": 0.6369519348144531, "cos_fa_negg": [0.8992404341697693], "loss": 0.6345866918563843}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.626319408416748, "diagnostics_wall_s": 0.048043251037597656, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.626319408416748, "evaluation_wall_s": 0.024016857147216797}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17310208, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s4.json b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s4.json new file mode 100644 index 0000000..e5465c9 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t1_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d1_t1_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7674304842948914}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.6115, "eval_loss": 0.6215746765136718, "wall_s": 2.742326021194458, "val_acc": 0.6115, "val_loss": 0.6215746765136718, "cos_fa_negg": [0.4044480323791504], "loss": 0.6178627014160156}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.661147356033325, "diagnostics_wall_s": 0.04787421226501465, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.661147356033325, "evaluation_wall_s": 0.02849102020263672}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17310208, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s0.json b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s0.json new file mode 100644 index 0000000..c2db09f --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d1_t2_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7000930905342102}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.535, "eval_loss": 0.6281340026855469, "wall_s": 2.7779202461242676, "val_acc": 0.535, "val_loss": 0.6281340026855469, "cos_fa_negg": [0.6817981004714966], "loss": 0.6318383812904358}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.7007174491882324, "diagnostics_wall_s": 0.04823732376098633, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.7007174491882324, "evaluation_wall_s": 0.02440619468688965}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17310208, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s1.json b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s1.json new file mode 100644 index 0000000..7819829 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d1_t2_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6932498812675476}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.502, "eval_loss": 0.6410772399902344, "wall_s": 2.677643299102783, "val_acc": 0.502, "val_loss": 0.6410772399902344, "cos_fa_negg": [0.7806419134140015], "loss": 0.643697202205658}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.602914810180664, "diagnostics_wall_s": 0.046007633209228516, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.602914810180664, "evaluation_wall_s": 0.02421116828918457}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17310208, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s2.json b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s2.json new file mode 100644 index 0000000..cb77f00 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d1_t2_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6975919604301453}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5365, "eval_loss": 0.6337516174316407, "wall_s": 2.6596624851226807, "val_acc": 0.5365, "val_loss": 0.6337516174316407, "cos_fa_negg": [0.9284895062446594], "loss": 0.6375528573989868}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.586501359939575, "diagnostics_wall_s": 0.04393434524536133, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.586501359939575, "evaluation_wall_s": 0.02458477020263672}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17310208, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s3.json b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s3.json new file mode 100644 index 0000000..079d143 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d1_t2_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7028841972351074}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5235, "eval_loss": 0.6399440002441407, "wall_s": 2.727836847305298, "val_acc": 0.5235, "val_loss": 0.6399440002441407, "cos_fa_negg": [0.898507297039032], "loss": 0.6430566906929016}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.6408398151397705, "diagnostics_wall_s": 0.05441713333129883, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.6408398151397705, "evaluation_wall_s": 0.027956008911132812}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17310208, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s4.json b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s4.json new file mode 100644 index 0000000..3483bfe --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d1_t2_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d1_t2_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.8385413289070129}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5305, "eval_loss": 0.6302141418457031, "wall_s": 2.6531147956848145, "val_acc": 0.5305, "val_loss": 0.6302141418457031, "cos_fa_negg": [0.8804345726966858], "loss": 0.6339011192321777}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 2.5646982192993164, "diagnostics_wall_s": 0.05951714515686035, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 2.5646982192993164, "evaluation_wall_s": 0.024212121963500977}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17310208, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s0.json b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s0.json new file mode 100644 index 0000000..ffdfa3f --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d4_t0_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6903697848320007}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.7025, "eval_loss": 0.4387386779785156, "wall_s": 6.018404245376587, "val_acc": 0.7025, "val_loss": 0.4387386779785156, "cos_fa_negg": [0.3807598948478699, 0.9523967504501343, 0.7027955055236816, 0.784724771976471], "loss": 0.44083914160728455, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.7468027472496033, 0.6991968750953674, 0.46745938062667847], "lesion_eval_acc": 0.7215, "lesion_eval_loss": 3.488897705078125, "lesion_acc_drop": -0.019000000000000017}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.894446611404419, "diagnostics_wall_s": 0.06199288368225098, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.894446611404419, "evaluation_wall_s": 0.05719876289367676}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17540608, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s1.json b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s1.json new file mode 100644 index 0000000..2b33f8e --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d4_t0_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.749412477016449}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.6975, "eval_loss": 0.6500516357421875, "wall_s": 5.704888343811035, "val_acc": 0.6975, "val_loss": 0.6500516357421875, "cos_fa_negg": [0.07989905774593353, -0.8221185803413391, 0.7079901099205017, 0.882909893989563], "loss": 0.6537783741950989, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.06225021928548813, 0.7829760909080505, 0.49877145886421204], "lesion_eval_acc": 0.5, "lesion_eval_loss": 0.8205497741699219, "lesion_acc_drop": 0.1975}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.578443765640259, "diagnostics_wall_s": 0.06435751914978027, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.578443765640259, "evaluation_wall_s": 0.05739331245422363}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17540608, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s2.json b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s2.json new file mode 100644 index 0000000..6cb6bd6 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d4_t0_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6906556487083435}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.957, "eval_loss": 0.3044500122070313, "wall_s": 5.546135902404785, "val_acc": 0.957, "val_loss": 0.3044500122070313, "cos_fa_negg": [0.7385568618774414, 0.9403408169746399, 0.4481012225151062, 0.8588445782661438], "loss": 0.3043282926082611, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.4007963240146637, 0.3674224317073822, 0.6594118475914001], "lesion_eval_acc": 0.582, "lesion_eval_loss": 2.79546826171875, "lesion_acc_drop": 0.375}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.413823127746582, "diagnostics_wall_s": 0.0741572380065918, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.413823127746582, "evaluation_wall_s": 0.053505897521972656}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17540608, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s3.json b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s3.json new file mode 100644 index 0000000..4fbaa51 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d4_t0_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7026023864746094}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.6835, "eval_loss": 0.56130859375, "wall_s": 5.565936088562012, "val_acc": 0.6835, "val_loss": 0.56130859375, "cos_fa_negg": [0.8604421615600586, 0.9097022414207458, -0.6087315678596497, 0.9837820529937744], "loss": 0.5715512633323669, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.6584221124649048, 0.2402328997850418, 0.47090792655944824], "lesion_eval_acc": 0.5, "lesion_eval_loss": 1.0498414916992187, "lesion_acc_drop": 0.1835}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.435111999511719, "diagnostics_wall_s": 0.06784248352050781, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.435111999511719, "evaluation_wall_s": 0.05846858024597168}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17540608, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s4.json b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s4.json new file mode 100644 index 0000000..ed14773 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t0_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d4_t0_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 1.1743755340576172}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.709, "eval_loss": 0.5532159729003906, "wall_s": 5.765451908111572, "val_acc": 0.709, "val_loss": 0.5532159729003906, "cos_fa_negg": [0.7983827590942383, 0.7034187316894531, 0.9918580055236816, 0.9242440462112427], "loss": 0.5628598928451538, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.5305265188217163, 0.10476234555244446, 0.2109975963830948], "lesion_eval_acc": 0.624, "lesion_eval_loss": 1.2968666381835938, "lesion_acc_drop": 0.08499999999999996}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.639735221862793, "diagnostics_wall_s": 0.06430649757385254, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.639735221862793, "evaluation_wall_s": 0.05666089057922363}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17540608, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s0.json b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s0.json new file mode 100644 index 0000000..162dca7 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d4_t1_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.686114490032196}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.809, "eval_loss": 0.47501167297363284, "wall_s": 6.049906253814697, "val_acc": 0.809, "val_loss": 0.47501167297363284, "cos_fa_negg": [0.2734767496585846, 0.9723815321922302, 0.781570315361023, 0.8252182602882385], "loss": 0.44219517707824707, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.7065876126289368, 0.6677806377410889, 0.4330540895462036], "lesion_eval_acc": 0.6815, "lesion_eval_loss": 3.6194608154296875, "lesion_acc_drop": 0.12750000000000006}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.924887657165527, "diagnostics_wall_s": 0.06189107894897461, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.924887657165527, "evaluation_wall_s": 0.05694270133972168}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17540608, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s1.json b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s1.json new file mode 100644 index 0000000..26fb36e --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d4_t1_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7337294220924377}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.602, "eval_loss": 0.6524281311035156, "wall_s": 5.764439582824707, "val_acc": 0.602, "val_loss": 0.6524281311035156, "cos_fa_negg": [0.09020137041807175, -0.7631667852401733, 0.7028098106384277, 0.8780344724655151], "loss": 0.6506175398826599, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.05208365246653557, 0.7791841626167297, 0.4998576045036316], "lesion_eval_acc": 0.5, "lesion_eval_loss": 0.8412230224609375, "lesion_acc_drop": 0.10199999999999998}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.643441200256348, "diagnostics_wall_s": 0.06030678749084473, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.643441200256348, "evaluation_wall_s": 0.05583357810974121}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17540608, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s2.json b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s2.json new file mode 100644 index 0000000..328f48b --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d4_t1_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6891603469848633}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.7345, "eval_loss": 0.5267286987304688, "wall_s": 7.112464666366577, "val_acc": 0.7345, "val_loss": 0.5267286987304688, "cos_fa_negg": [0.8189562559127808, -0.4384288191795349, 0.7410873174667358, 0.6817660927772522], "loss": 0.5338773727416992, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.09999122470617294, 0.4018876254558563, 0.7487749457359314], "lesion_eval_acc": 0.5495, "lesion_eval_loss": 2.036950988769531, "lesion_acc_drop": 0.18500000000000005}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 6.988081693649292, "diagnostics_wall_s": 0.06311249732971191, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 6.988081693649292, "evaluation_wall_s": 0.05582165718078613}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17540608, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s3.json b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s3.json new file mode 100644 index 0000000..fa0681a --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d4_t1_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7055310606956482}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.9675, "eval_loss": 0.18122200012207032, "wall_s": 5.670015811920166, "val_acc": 0.9675, "val_loss": 0.18122200012207032, "cos_fa_negg": [0.6549974679946899, 0.9112385511398315, 0.8708909153938293, 0.8902857303619385], "loss": 0.17875875532627106, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.5672092437744141, 0.14307843148708344, 0.3526569604873657], "lesion_eval_acc": 0.6875, "lesion_eval_loss": 0.5926900634765625, "lesion_acc_drop": 0.28}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.542003393173218, "diagnostics_wall_s": 0.06803250312805176, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.542003393173218, "evaluation_wall_s": 0.0552973747253418}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17540608, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s4.json b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s4.json new file mode 100644 index 0000000..a1a48a4 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t1_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d4_t1_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 1.2137476205825806}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.953, "eval_loss": 0.30602383422851565, "wall_s": 5.9932825565338135, "val_acc": 0.953, "val_loss": 0.30602383422851565, "cos_fa_negg": [0.692764401435852, 0.6571060419082642, 0.9609348773956299, 0.9080684781074524], "loss": 0.29758796095848083, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.581433117389679, 0.23921656608581543, 0.1754363626241684], "lesion_eval_acc": 0.5865, "lesion_eval_loss": 1.8497723999023437, "lesion_acc_drop": 0.36649999999999994}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.861577987670898, "diagnostics_wall_s": 0.06736254692077637, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.861577987670898, "evaluation_wall_s": 0.05954861640930176}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17540608, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s0.json b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s0.json new file mode 100644 index 0000000..391f3a6 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d4_t2_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7004839181900024}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.7315, "eval_loss": 0.4644243927001953, "wall_s": 5.507065534591675, "val_acc": 0.7315, "val_loss": 0.4644243927001953, "cos_fa_negg": [0.26633673906326294, 0.9680317640304565, 0.7551535367965698, 0.8186001777648926], "loss": 0.4706776440143585, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.7198461890220642, 0.6803547739982605, 0.45424333214759827], "lesion_eval_acc": 0.6835, "lesion_eval_loss": 3.984370849609375, "lesion_acc_drop": 0.04800000000000004}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.384226083755493, "diagnostics_wall_s": 0.06237363815307617, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.384226083755493, "evaluation_wall_s": 0.055866241455078125}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17540608, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s1.json b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s1.json new file mode 100644 index 0000000..5ec9a07 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d4_t2_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7904043793678284}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.464, "eval_loss": 0.6586670532226563, "wall_s": 6.05906343460083, "val_acc": 0.464, "val_loss": 0.6586670532226563, "cos_fa_negg": [0.07074272632598877, -0.7343921661376953, 0.6853258609771729, 0.8569676280021667], "loss": 0.6590549945831299, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.05030565336346626, 0.7675413489341736, 0.5057328939437866], "lesion_eval_acc": 0.5, "lesion_eval_loss": 0.8384515686035157, "lesion_acc_drop": -0.035999999999999976}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.934049606323242, "diagnostics_wall_s": 0.06332159042358398, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.934049606323242, "evaluation_wall_s": 0.05627703666687012}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17540608, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s2.json b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s2.json new file mode 100644 index 0000000..4fdee8c --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d4_t2_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6941237449645996}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.9545, "eval_loss": 0.25613199615478516, "wall_s": 5.929734468460083, "val_acc": 0.9545, "val_loss": 0.25613199615478516, "cos_fa_negg": [0.8122324347496033, 0.9910876750946045, 0.6968773603439331, 0.6953873634338379], "loss": 0.26535528898239136, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.28186851739883423, 0.3978812098503113, 0.5451219081878662], "lesion_eval_acc": 0.581, "lesion_eval_loss": 2.691932861328125, "lesion_acc_drop": 0.37350000000000005}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.802331447601318, "diagnostics_wall_s": 0.06623482704162598, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.802331447601318, "evaluation_wall_s": 0.05598163604736328}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17540608, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s3.json b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s3.json new file mode 100644 index 0000000..c74ec4a --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d4_t2_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.696068286895752}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.992, "eval_loss": 0.1585829620361328, "wall_s": 5.6263322830200195, "val_acc": 0.992, "val_loss": 0.1585829620361328, "cos_fa_negg": [0.8624885082244873, 0.9677263498306274, 0.45145902037620544, 0.9651761054992676], "loss": 0.1604032963514328, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.6198669672012329, 0.30877771973609924, 0.4941253364086151], "lesion_eval_acc": 0.73, "lesion_eval_loss": 0.4453875732421875, "lesion_acc_drop": 0.262}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.500669240951538, "diagnostics_wall_s": 0.06495523452758789, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.500669240951538, "evaluation_wall_s": 0.05588245391845703}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17540608, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s4.json b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s4.json new file mode 100644 index 0000000..9ce95a4 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_fa_d4_t2_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "fa", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_fa_d4_t2_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 1.1055001020431519}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.6145, "eval_loss": 0.570775634765625, "wall_s": 5.556745767593384, "val_acc": 0.6145, "val_loss": 0.570775634765625, "cos_fa_negg": [0.7983068823814392, 0.6619317531585693, 0.960556149482727, 0.9008275866508484], "loss": 0.5740307569503784, "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.5150701999664307, 0.09795241057872772, 0.21567916870117188], "lesion_eval_acc": 0.617, "lesion_eval_loss": 1.265248779296875, "lesion_acc_drop": -0.0024999999999999467}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.4379048347473145, "diagnostics_wall_s": 0.06064653396606445, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.4379048347473145, "evaluation_wall_s": 0.05368518829345703}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 0, "calibration_batch_loss_evaluations": 0, "calibration_example_loss_evaluations": 0, "calibration_forward_equivalent_examples": 0.0, "training_forward_equivalent_examples": 640000.0, "max_perturbation_batch_expansion": 0}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17540608, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s0.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s0.json new file mode 100644 index 0000000..c6a03cb --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7099499106407166}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.7135, "eval_loss": 0.6042601318359375, "wall_s": 3.8217966556549072, "val_acc": 0.7135, "val_loss": 0.6042601318359375, "cos_r_negg": [0.9998776316642761], "cos_innovation_negg": [0.9998776316642761], "cos_apical_negg": [0.9998776316642761], "cos_Ac_negg": [0.9998776316642761], "r_norm": [3.2442517280578613], "g_norm": [0.007123492658138275], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 3.7473464012145996, "diagnostics_wall_s": 0.04571270942687988, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 3.7473464012145996, "evaluation_wall_s": 0.024243593215942383}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s1.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s1.json new file mode 100644 index 0000000..5078a08 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6922826766967773}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.6315, "eval_loss": 0.5984332885742187, "wall_s": 4.2087414264678955, "val_acc": 0.6315, "val_loss": 0.5984332885742187, "cos_r_negg": [0.999922513961792], "cos_innovation_negg": [0.999922513961792], "cos_apical_negg": [0.999922513961792], "cos_Ac_negg": [0.999922513961792], "r_norm": [2.8550167083740234], "g_norm": [0.006600195541977882], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 4.128992795944214, "diagnostics_wall_s": 0.04770207405090332, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 4.128992795944214, "evaluation_wall_s": 0.027198076248168945}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s2.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s2.json new file mode 100644 index 0000000..50db2d0 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7055632472038269}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5935, "eval_loss": 0.5991917114257812, "wall_s": 4.107821941375732, "val_acc": 0.5935, "val_loss": 0.5991917114257812, "cos_r_negg": [0.9998976588249207], "cos_innovation_negg": [0.9998976588249207], "cos_apical_negg": [0.9998976588249207], "cos_Ac_negg": [0.9998976588249207], "r_norm": [2.836979389190674], "g_norm": [0.006569740828126669], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 4.028418779373169, "diagnostics_wall_s": 0.04642653465270996, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 4.028418779373169, "evaluation_wall_s": 0.02491927146911621}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s3.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s3.json new file mode 100644 index 0000000..6ebbe06 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7144617438316345}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.558, "eval_loss": 0.6029836120605468, "wall_s": 4.1979968547821045, "val_acc": 0.558, "val_loss": 0.6029836120605468, "cos_r_negg": [0.9998957514762878], "cos_innovation_negg": [0.9998957514762878], "cos_apical_negg": [0.9998957514762878], "cos_Ac_negg": [0.9998957514762878], "r_norm": [2.668304920196533], "g_norm": [0.006326030008494854], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 4.120422601699829, "diagnostics_wall_s": 0.04749274253845215, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 4.120422601699829, "evaluation_wall_s": 0.02514815330505371}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s4.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s4.json new file mode 100644 index 0000000..9f3bc8d --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d1_t0_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7867429256439209}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.7295, "eval_loss": 0.517715087890625, "wall_s": 4.068825006484985, "val_acc": 0.7295, "val_loss": 0.517715087890625, "cos_r_negg": [0.9999740719795227], "cos_innovation_negg": [0.9999740719795227], "cos_apical_negg": [0.9999740719795227], "cos_Ac_negg": [0.9999740719795227], "r_norm": [3.901029586791992], "g_norm": [0.008658962324261665], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 3.986222267150879, "diagnostics_wall_s": 0.0530397891998291, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 3.986222267150879, "evaluation_wall_s": 0.02480483055114746}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s0.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s0.json new file mode 100644 index 0000000..9ed577b --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7144274711608887}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5435, "eval_loss": 0.6050716247558594, "wall_s": 4.012980222702026, "val_acc": 0.5435, "val_loss": 0.6050716247558594, "cos_r_negg": [0.9997878670692444], "cos_innovation_negg": [0.9997878670692444], "cos_apical_negg": [0.9997878670692444], "cos_Ac_negg": [0.9997878670692444], "r_norm": [3.2369086742401123], "g_norm": [0.007149500772356987], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 3.9372355937957764, "diagnostics_wall_s": 0.045932769775390625, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 3.9372355937957764, "evaluation_wall_s": 0.024921417236328125}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s1.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s1.json new file mode 100644 index 0000000..f998cca --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.691861629486084}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.6065, "eval_loss": 0.6069668273925781, "wall_s": 4.294641494750977, "val_acc": 0.6065, "val_loss": 0.6069668273925781, "cos_r_negg": [0.9999315738677979], "cos_innovation_negg": [0.9999315738677979], "cos_apical_negg": [0.9999315738677979], "cos_Ac_negg": [0.9999315738677979], "r_norm": [2.844742774963379], "g_norm": [0.006638430058956146], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 4.215763807296753, "diagnostics_wall_s": 0.04869484901428223, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 4.215763807296753, "evaluation_wall_s": 0.024796247482299805}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s2.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s2.json new file mode 100644 index 0000000..7e7baf5 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7091569304466248}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.7295, "eval_loss": 0.619099365234375, "wall_s": 4.391073703765869, "val_acc": 0.7295, "val_loss": 0.619099365234375, "cos_r_negg": [0.999864399433136], "cos_innovation_negg": [0.999864399433136], "cos_apical_negg": [0.999864399433136], "cos_Ac_negg": [0.999864399433136], "r_norm": [2.648947238922119], "g_norm": [0.006350507959723473], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 4.315802335739136, "diagnostics_wall_s": 0.04552912712097168, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 4.315802335739136, "evaluation_wall_s": 0.02494645118713379}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s3.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s3.json new file mode 100644 index 0000000..f6b5c87 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7197580337524414}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.7095, "eval_loss": 0.6599753723144531, "wall_s": 4.307591199874878, "val_acc": 0.7095, "val_loss": 0.6599753723144531, "cos_r_negg": [0.9940416812896729], "cos_innovation_negg": [0.9940416812896729], "cos_apical_negg": [0.9940416812896729], "cos_Ac_negg": [0.9940416812896729], "r_norm": [1.1960170269012451], "g_norm": [0.003565517719835043], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 4.2271568775177, "diagnostics_wall_s": 0.047910213470458984, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 4.2271568775177, "evaluation_wall_s": 0.027333498001098633}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s4.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s4.json new file mode 100644 index 0000000..5f44b3f --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d1_t1_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7674305438995361}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.7295, "eval_loss": 0.5247836608886719, "wall_s": 3.9241554737091064, "val_acc": 0.7295, "val_loss": 0.5247836608886719, "cos_r_negg": [0.9999526739120483], "cos_innovation_negg": [0.9999526739120483], "cos_apical_negg": [0.9999526739120483], "cos_Ac_negg": [0.9999526739120483], "r_norm": [3.909113645553589], "g_norm": [0.008722091093659401], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 3.839465379714966, "diagnostics_wall_s": 0.04948544502258301, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 3.839465379714966, "evaluation_wall_s": 0.024755239486694336}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s0.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s0.json new file mode 100644 index 0000000..b79d9fc --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.700093150138855}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.6385, "eval_loss": 0.6210680541992187, "wall_s": 4.071274757385254, "val_acc": 0.6385, "val_loss": 0.6210680541992187, "cos_r_negg": [0.9997798800468445], "cos_innovation_negg": [0.9997798800468445], "cos_apical_negg": [0.9997798800468445], "cos_Ac_negg": [0.9997798800468445], "r_norm": [3.254207134246826], "g_norm": [0.007089059334248304], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 3.9992709159851074, "diagnostics_wall_s": 0.04352617263793945, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 3.9992709159851074, "evaluation_wall_s": 0.024050474166870117}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s1.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s1.json new file mode 100644 index 0000000..4483c30 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6932498216629028}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.613, "eval_loss": 0.6148249816894531, "wall_s": 4.006633043289185, "val_acc": 0.613, "val_loss": 0.6148249816894531, "cos_r_negg": [0.9998847842216492], "cos_innovation_negg": [0.9998847842216492], "cos_apical_negg": [0.9998847842216492], "cos_Ac_negg": [0.9998847842216492], "r_norm": [2.8224780559539795], "g_norm": [0.006553794257342815], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 3.92795467376709, "diagnostics_wall_s": 0.049527645111083984, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 3.92795467376709, "evaluation_wall_s": 0.02442789077758789}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s2.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s2.json new file mode 100644 index 0000000..aa6e3d4 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6975919008255005}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.6355, "eval_loss": 0.6171273193359375, "wall_s": 4.152236223220825, "val_acc": 0.6355, "val_loss": 0.6171273193359375, "cos_r_negg": [0.999872088432312], "cos_innovation_negg": [0.999872088432312], "cos_apical_negg": [0.999872088432312], "cos_Ac_negg": [0.999872088432312], "r_norm": [2.7898759841918945], "g_norm": [0.006504997611045837], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 4.0797271728515625, "diagnostics_wall_s": 0.04322361946105957, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 4.0797271728515625, "evaluation_wall_s": 0.024546384811401367}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s3.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s3.json new file mode 100644 index 0000000..ec85fb6 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7028841376304626}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.6425, "eval_loss": 0.6185914306640625, "wall_s": 4.137370824813843, "val_acc": 0.6425, "val_loss": 0.6185914306640625, "cos_r_negg": [0.9998453855514526], "cos_innovation_negg": [0.9998453855514526], "cos_apical_negg": [0.9998453855514526], "cos_Ac_negg": [0.9998453855514526], "r_norm": [2.6966309547424316], "g_norm": [0.006380176171660423], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 4.056280136108398, "diagnostics_wall_s": 0.049558401107788086, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 4.056280136108398, "evaluation_wall_s": 0.026686429977416992}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s4.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s4.json new file mode 100644 index 0000000..9e6f233 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d1_t2_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.8385413885116577}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.728, "eval_loss": 0.5332000122070313, "wall_s": 3.989596366882324, "val_acc": 0.728, "val_loss": 0.5332000122070313, "cos_r_negg": [0.9999899864196777], "cos_innovation_negg": [0.9999899864196777], "cos_apical_negg": [0.9999899864196777], "cos_Ac_negg": [0.9999899864196777], "r_norm": [3.8140172958374023], "g_norm": [0.008471222594380379], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 3.9157958030700684, "diagnostics_wall_s": 0.0434117317199707, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 3.9157958030700684, "evaluation_wall_s": 0.026030302047729492}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396224, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s0.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s0.json new file mode 100644 index 0000000..b5c154a --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6903698444366455}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.943, "eval_loss": 0.14808297729492187, "wall_s": 10.704808712005615, "val_acc": 0.943, "val_loss": 0.14808297729492187, "cos_r_negg": [0.4984794855117798, 0.7471765279769897, 0.9844636917114258, 0.9806895852088928], "cos_innovation_negg": [0.4984794855117798, 0.7471765279769897, 0.9844636917114258, 0.9806895852088928], "cos_apical_negg": [0.4984794855117798, 0.7471765279769897, 0.9844636917114258, 0.9806895852088928], "cos_Ac_negg": [0.4984794855117798, 0.7471765279769897, 0.9844636917114258, 0.9806895852088928], "r_norm": [1.3431203365325928, 1.4957077503204346, 1.0617897510528564, 0.8341032266616821], "g_norm": [0.0054263281635940075, 0.004377680830657482, 0.002934478921815753, 0.002219215501099825], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [1.455359935760498, 0.2696658968925476, 0.5878478288650513], "lesion_eval_acc": 0.75, "lesion_eval_loss": 0.3969853210449219, "lesion_acc_drop": 0.19299999999999995}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 10.585901021957397, "diagnostics_wall_s": 0.07331490516662598, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 10.585901021957397, "evaluation_wall_s": 0.040700435638427734}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s1.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s1.json new file mode 100644 index 0000000..fae7b35 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7494125366210938}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.718, "eval_loss": 0.5313345947265625, "wall_s": 10.41555643081665, "val_acc": 0.718, "val_loss": 0.5313345947265625, "cos_r_negg": [0.9837238788604736, 0.9956188201904297, 0.9917439818382263, 0.9767022132873535], "cos_innovation_negg": [0.9837238788604736, 0.9956188201904297, 0.9917439818382263, 0.9767022132873535], "cos_apical_negg": [0.9837238788604736, 0.9956188201904297, 0.9917439818382263, 0.9767022132873535], "cos_Ac_negg": [0.9837238788604736, 0.9956188201904297, 0.9917439818382263, 0.9767022132873535], "r_norm": [3.920393466949463, 3.5351293087005615, 2.2725727558135986, 2.2366135120391846], "g_norm": [0.00900379940867424, 0.009184032678604126, 0.005431623198091984, 0.0052651213482022285], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.42316144704818726, 0.45310166478157043, 0.7213063836097717], "lesion_eval_acc": 0.5, "lesion_eval_loss": 0.9191639709472657, "lesion_acc_drop": 0.21799999999999997}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 10.300774812698364, "diagnostics_wall_s": 0.06851339340209961, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 10.300774812698364, "evaluation_wall_s": 0.04140520095825195}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s2.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s2.json new file mode 100644 index 0000000..c0d65a1 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6906555891036987}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.7645, "eval_loss": 0.4268807067871094, "wall_s": 10.210656881332397, "val_acc": 0.7645, "val_loss": 0.4268807067871094, "cos_r_negg": [0.3803340196609497, 0.3853844106197357, 0.8091791868209839, 0.99535071849823], "cos_innovation_negg": [0.3803340196609497, 0.3853844106197357, 0.8091791868209839, 0.99535071849823], "cos_apical_negg": [0.3803340196609497, 0.3853844106197357, 0.8091791868209839, 0.99535071849823], "cos_Ac_negg": [0.3803340196609497, 0.3853844106197357, 0.8091791868209839, 0.99535071849823], "r_norm": [3.798327684402466, 3.8802895545959473, 3.7749059200286865, 1.9753646850585938], "g_norm": [0.011167047545313835, 0.010896231979131699, 0.009289173409342766, 0.004587862640619278], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.4966542720794678, 0.7198088765144348, 1.258798599243164], "lesion_eval_acc": 0.5935, "lesion_eval_loss": 3.8238804931640624, "lesion_acc_drop": 0.17099999999999993}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 10.093049764633179, "diagnostics_wall_s": 0.07129287719726562, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 10.093049764633179, "evaluation_wall_s": 0.04147839546203613}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s3.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s3.json new file mode 100644 index 0000000..1a404fb --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7026023268699646}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5835, "eval_loss": 0.5610132141113281, "wall_s": 10.165930986404419, "val_acc": 0.5835, "val_loss": 0.5610132141113281, "cos_r_negg": [0.8743358850479126, 0.940854012966156, 0.9968980550765991, 0.9943938851356506], "cos_innovation_negg": [0.8743358850479126, 0.940854012966156, 0.9968980550765991, 0.9943938851356506], "cos_apical_negg": [0.8743358850479126, 0.940854012966156, 0.9968980550765991, 0.9943938851356506], "cos_Ac_negg": [0.8743358850479126, 0.940854012966156, 0.9968980550765991, 0.9943938851356506], "r_norm": [2.66884708404541, 3.1014907360076904, 3.0315423011779785, 2.5119330883026123], "g_norm": [0.005882163532078266, 0.006724152714014053, 0.006184111349284649, 0.005286586470901966], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.25876060128211975, 0.07180501520633698, 0.6527635455131531], "lesion_eval_acc": 0.5, "lesion_eval_loss": 0.6458817443847656, "lesion_acc_drop": 0.08350000000000002}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 10.049697637557983, "diagnostics_wall_s": 0.07040619850158691, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 10.049697637557983, "evaluation_wall_s": 0.04107069969177246}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s4.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s4.json new file mode 100644 index 0000000..a45fcd1 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d4_t0_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 1.1743755340576172}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.625, "eval_loss": 0.5968040161132813, "wall_s": 10.755664825439453, "val_acc": 0.625, "val_loss": 0.5968040161132813, "cos_r_negg": [0.9945129752159119, 0.9705206155776978, 0.9994956851005554, 0.999819815158844], "cos_innovation_negg": [0.9945129752159119, 0.9705206155776978, 0.9994956851005554, 0.999819815158844], "cos_apical_negg": [0.9945129752159119, 0.9705206155776978, 0.9994956851005554, 0.999819815158844], "cos_Ac_negg": [0.9945129752159119, 0.9705206155776978, 0.9994956851005554, 0.999819815158844], "r_norm": [3.143918037414551, 2.595519542694092, 2.3923661708831787, 2.3235230445861816], "g_norm": [0.006242428440600634, 0.005687946453690529, 0.005275203846395016, 0.004979344084858894], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.5525150299072266, 0.38599294424057007, 0.18380124866962433], "lesion_eval_acc": 0.603, "lesion_eval_loss": 0.6570440368652344, "lesion_acc_drop": 0.02200000000000002}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 10.635491847991943, "diagnostics_wall_s": 0.07047629356384277, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 10.635491847991943, "evaluation_wall_s": 0.044860124588012695}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s0.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s0.json new file mode 100644 index 0000000..ebf51b5 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.686114490032196}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.98, "eval_loss": 0.08475372695922852, "wall_s": 10.560627460479736, "val_acc": 0.98, "val_loss": 0.08475372695922852, "cos_r_negg": [0.8068772554397583, 0.9649519920349121, 0.9943944215774536, 0.9917788505554199], "cos_innovation_negg": [0.8068772554397583, 0.9649519920349121, 0.9943944215774536, 0.9917788505554199], "cos_apical_negg": [0.8068772554397583, 0.9649519920349121, 0.9943944215774536, 0.9917788505554199], "cos_Ac_negg": [0.8068772554397583, 0.9649519920349121, 0.9943944215774536, 0.9917788505554199], "r_norm": [1.7875993251800537, 1.5450325012207031, 1.081005334854126, 0.5267228484153748], "g_norm": [0.011391619220376015, 0.006912127137184143, 0.005373181309551001, 0.0017763726646080613], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [1.27436363697052, 0.17199987173080444, 0.6046966314315796], "lesion_eval_acc": 0.5, "lesion_eval_loss": 1.284403564453125, "lesion_acc_drop": 0.48}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 10.430111646652222, "diagnostics_wall_s": 0.08903384208679199, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 10.430111646652222, "evaluation_wall_s": 0.03662610054016113}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s1.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s1.json new file mode 100644 index 0000000..65dba87 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7337294816970825}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5415, "eval_loss": 0.5803764038085938, "wall_s": 9.850732803344727, "val_acc": 0.5415, "val_loss": 0.5803764038085938, "cos_r_negg": [0.9904084205627441, 0.9376134276390076, 0.9324054718017578, 0.9886455535888672], "cos_innovation_negg": [0.9904084205627441, 0.9376134276390076, 0.9324054718017578, 0.9886455535888672], "cos_apical_negg": [0.9904084205627441, 0.9376134276390076, 0.9324054718017578, 0.9886455535888672], "cos_Ac_negg": [0.9904084205627441, 0.9376134276390076, 0.9324054718017578, 0.9886455535888672], "r_norm": [3.6325924396514893, 3.0356807708740234, 2.5806682109832764, 2.4982333183288574], "g_norm": [0.008355510421097279, 0.007141077890992165, 0.005960159469395876, 0.00567194540053606], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.24353612959384918, 0.48792389035224915, 0.7457113862037659], "lesion_eval_acc": 0.669, "lesion_eval_loss": 0.8552682495117188, "lesion_acc_drop": -0.12750000000000006}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 9.73811936378479, "diagnostics_wall_s": 0.06741547584533691, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 9.73811936378479, "evaluation_wall_s": 0.040467262268066406}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s2.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s2.json new file mode 100644 index 0000000..3399722 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6891602873802185}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.834, "eval_loss": 0.43524066162109376, "wall_s": 10.593032121658325, "val_acc": 0.834, "val_loss": 0.43524066162109376, "cos_r_negg": [0.6454845666885376, 0.7907583117485046, 0.9466684460639954, 0.9947988986968994], "cos_innovation_negg": [0.6454845666885376, 0.7907583117485046, 0.9466684460639954, 0.9947988986968994], "cos_apical_negg": [0.6454845666885376, 0.7907583117485046, 0.9466684460639954, 0.9947988986968994], "cos_Ac_negg": [0.6454845666885376, 0.7907583117485046, 0.9466684460639954, 0.9947988986968994], "r_norm": [3.5640435218811035, 3.6834654808044434, 3.6468286514282227, 1.9396458864212036], "g_norm": [0.011073324829339981, 0.011038878932595253, 0.009526113048195839, 0.004571080207824707], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.5253376364707947, 0.7072277665138245, 1.2511272430419922], "lesion_eval_acc": 0.5945, "lesion_eval_loss": 4.649655029296875, "lesion_acc_drop": 0.23949999999999994}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 10.472795724868774, "diagnostics_wall_s": 0.07233953475952148, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 10.472795724868774, "evaluation_wall_s": 0.04293179512023926}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s3.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s3.json new file mode 100644 index 0000000..9b95a6c --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.705531120300293}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.904, "eval_loss": 0.1922312240600586, "wall_s": 13.10716700553894, "val_acc": 0.904, "val_loss": 0.1922312240600586, "cos_r_negg": [0.8487608432769775, 0.9046051502227783, 0.9973889589309692, 0.9829939603805542], "cos_innovation_negg": [0.8487608432769775, 0.9046051502227783, 0.9973889589309692, 0.9829939603805542], "cos_apical_negg": [0.8487608432769775, 0.9046051502227783, 0.9973889589309692, 0.9829939603805542], "cos_Ac_negg": [0.8487608432769775, 0.9046051502227783, 0.9973889589309692, 0.9829939603805542], "r_norm": [1.9577268362045288, 1.5276708602905273, 1.4349093437194824, 0.7424835562705994], "g_norm": [0.01130524929612875, 0.008212299086153507, 0.006950871087610722, 0.0024630320258438587], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.6706282496452332, 0.4811720550060272, 0.3360276520252228], "lesion_eval_acc": 0.5605, "lesion_eval_loss": 1.260206787109375, "lesion_acc_drop": 0.3435}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 12.983716011047363, "diagnostics_wall_s": 0.07724666595458984, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 12.983716011047363, "evaluation_wall_s": 0.04076385498046875}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s4.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s4.json new file mode 100644 index 0000000..299d68d --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 1, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d4_t1_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 1, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "2eaf578917eb435bde86ba85631fa86644d2621c9be3ae6754d8128180c4b302", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 1.213747501373291}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.556, "eval_loss": 0.5772188415527344, "wall_s": 10.411751747131348, "val_acc": 0.556, "val_loss": 0.5772188415527344, "cos_r_negg": [0.9878685474395752, 0.9620581269264221, 0.9984392523765564, 0.9998359084129333], "cos_innovation_negg": [0.9878685474395752, 0.9620581269264221, 0.9984392523765564, 0.9998359084129333], "cos_apical_negg": [0.9878685474395752, 0.9620581269264221, 0.9984392523765564, 0.9998359084129333], "cos_Ac_negg": [0.9878685474395752, 0.9620581269264221, 0.9984392523765564, 0.9998359084129333], "r_norm": [2.91984224319458, 2.4763145446777344, 2.301741123199463, 2.2300872802734375], "g_norm": [0.005881824530661106, 0.005384676158428192, 0.0051411837339401245, 0.0049733966588974], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.4328557252883911, 0.38256096839904785, 0.14993762969970703], "lesion_eval_acc": 0.484, "lesion_eval_loss": 0.6305642395019532, "lesion_acc_drop": 0.07200000000000006}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 10.299676179885864, "diagnostics_wall_s": 0.06645703315734863, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 10.299676179885864, "evaluation_wall_s": 0.04036760330200195}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s0.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s0.json new file mode 100644 index 0000000..423dce3 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7004837989807129}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5, "eval_loss": 13480.0225, "wall_s": 10.444090366363525, "val_acc": 0.5, "val_loss": 13480.0225, "cos_r_negg": [NaN, -1.0, NaN, NaN], "cos_innovation_negg": [NaN, -1.0, NaN, NaN], "cos_apical_negg": [NaN, -1.0, NaN, NaN], "cos_Ac_negg": [NaN, -1.0, NaN, NaN], "r_norm": [159500.53125, 161555.28125, 173405.640625, 155405.265625], "g_norm": [916.73876953125, 916.71142578125, 916.71142578125, 916.71142578125], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [1280851321028608.0, 0.0, 0.0], "lesion_eval_acc": 0.5, "lesion_eval_loss": 13480.0225, "lesion_acc_drop": 0.0}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 10.321823358535767, "diagnostics_wall_s": 0.07626557350158691, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 10.321823358535767, "evaluation_wall_s": 0.04099011421203613}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s1.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s1.json new file mode 100644 index 0000000..689b79e --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7904044985771179}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.6475, "eval_loss": 0.606488525390625, "wall_s": 10.647256851196289, "val_acc": 0.6475, "val_loss": 0.606488525390625, "cos_r_negg": [0.987451434135437, 0.9417468309402466, 0.933337926864624, 0.981619656085968], "cos_innovation_negg": [0.987451434135437, 0.9417468309402466, 0.933337926864624, 0.981619656085968], "cos_apical_negg": [0.987451434135437, 0.9417468309402466, 0.933337926864624, 0.981619656085968], "cos_Ac_negg": [0.987451434135437, 0.9417468309402466, 0.933337926864624, 0.981619656085968], "r_norm": [3.610081195831299, 2.9995689392089844, 2.6030001640319824, 2.4571006298065186], "g_norm": [0.007628738880157471, 0.006662414409220219, 0.005830155685544014, 0.005557000637054443], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.2766271233558655, 0.4847310185432434, 0.719550609588623], "lesion_eval_acc": 0.706, "lesion_eval_loss": 0.7659820556640625, "lesion_acc_drop": -0.058499999999999996}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 10.530057191848755, "diagnostics_wall_s": 0.07096266746520996, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 10.530057191848755, "evaluation_wall_s": 0.041341543197631836}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s2.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s2.json new file mode 100644 index 0000000..38a1e1e --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6941238045692444}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.822, "eval_loss": 0.4962158203125, "wall_s": 10.018813133239746, "val_acc": 0.822, "val_loss": 0.4962158203125, "cos_r_negg": [0.7274631261825562, -0.0664236769080162, 0.969781756401062, 0.9946377277374268], "cos_innovation_negg": [0.7274631261825562, -0.0664236769080162, 0.969781756401062, 0.9946377277374268], "cos_apical_negg": [0.7274631261825562, -0.0664236769080162, 0.969781756401062, 0.9946377277374268], "cos_Ac_negg": [0.7274631261825562, -0.0664236769080162, 0.969781756401062, 0.9946377277374268], "r_norm": [4.307065963745117, 4.266366958618164, 4.113178253173828, 2.2948708534240723], "g_norm": [0.011777931824326515, 0.011061581782996655, 0.009688959456980228, 0.005156431347131729], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.3406045734882355, 0.5751078128814697, 0.6724721789360046], "lesion_eval_acc": 0.5965, "lesion_eval_loss": 2.0259490966796876, "lesion_acc_drop": 0.22549999999999992}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 9.910342693328857, "diagnostics_wall_s": 0.06596159934997559, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 9.910342693328857, "evaluation_wall_s": 0.03792428970336914}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s3.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s3.json new file mode 100644 index 0000000..f525fa9 --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s3.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s3", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6960683465003967}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.706, "eval_loss": 0.6010469665527344, "wall_s": 10.204753637313843, "val_acc": 0.706, "val_loss": 0.6010469665527344, "cos_r_negg": [0.9182963371276855, 0.9569817185401917, 0.9331937432289124, 0.9882691502571106], "cos_innovation_negg": [0.9182963371276855, 0.9569817185401917, 0.9331937432289124, 0.9882691502571106], "cos_apical_negg": [0.9182963371276855, 0.9569817185401917, 0.9331937432289124, 0.9882691502571106], "cos_Ac_negg": [0.9182963371276855, 0.9569817185401917, 0.9331937432289124, 0.9882691502571106], "r_norm": [3.3123316764831543, 3.2478415966033936, 3.0539560317993164, 2.598323345184326], "g_norm": [0.007931765168905258, 0.007240200415253639, 0.006589421071112156, 0.005645371042191982], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.49712440371513367, 0.2127549946308136, 0.5075206756591797], "lesion_eval_acc": 0.7115, "lesion_eval_loss": 0.6108424072265625, "lesion_acc_drop": -0.00550000000000006}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 10.089348793029785, "diagnostics_wall_s": 0.06796526908874512, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 10.089348793029785, "evaluation_wall_s": 0.0415959358215332}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s4.json b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s4.json new file mode 100644 index 0000000..52da1cb --- /dev/null +++ b/results/c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s4.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 2, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_local_val_v1_tent_l2_w8_sdil_d4_t2_s4", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 2, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "5c0cebb2220d23e8aab351ee7a5a8d97c1c8d8eee604e8e1aac6d8d8108629e7", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "51301f6738a8feef6830a4ce30e09428e619abd7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 1.1055001020431519}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.613, "eval_loss": 0.6409906616210937, "wall_s": 10.187198877334595, "val_acc": 0.613, "val_loss": 0.6409906616210937, "cos_r_negg": [0.974547266960144, 0.9372525215148926, 0.9995588660240173, -0.703455924987793], "cos_innovation_negg": [0.974547266960144, 0.9372525215148926, 0.9995588660240173, -0.703455924987793], "cos_apical_negg": [0.974547266960144, 0.9372525215148926, 0.9995588660240173, -0.703455924987793], "cos_Ac_negg": [0.974547266960144, 0.9372525215148926, 0.9995588660240173, -0.703455924987793], "r_norm": [2.8922979831695557, 2.5213658809661865, 2.395078659057617, 2.301630973815918], "g_norm": [0.005787475034594536, 0.005541402846574783, 0.005268101580440998, 0.005023573059588671], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.4897635579109192, 0.4147346019744873, 0.14394301176071167], "lesion_eval_acc": 0.5515, "lesion_eval_loss": 0.7152392883300781, "lesion_acc_drop": 0.0615}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 10.070974588394165, "diagnostics_wall_s": 0.07088255882263184, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 10.070974588394165, "evaluation_wall_s": 0.040782928466796875}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 1280, "calibration_example_loss_evaluations": 327680, "calibration_forward_equivalent_examples": 491520.0, "training_forward_equivalent_examples": 1131520.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17558016, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file -- cgit v1.2.3