From cbecb171b1af77fe6510fd60f1dfa4c7938a20d6 Mon Sep 17 00:00:00 2001 From: Yuren Hao Date: Sun, 5 Jul 2026 03:53:57 -0500 Subject: =?UTF-8?q?holofast:=20exact=20halved-jvp=20AEP=20track=20?= =?UTF-8?q?=E2=80=94=201.55x=20nudged=20phase,=20gradient-gate=20passed?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit holo_a_track computed the doubled-batch jvp/vjp on [v0; -v0] at a shared anchor zbar — exact antisymmetric redundancy (phase deviations from the common mode are exact negatives). holo_a_track_fast computes at batch B and mirrors: single-eval parity 6e-7 (exact); trajectory-level 45% divergence SHARED with the original's own FD noise floor (1e-6 state noise -> 49% self-divergence — the 2r=0.04 finite difference amplifies fp noise; training averages it via pema/momentum). Ship gate: cos(EP,BPTT) orig vs fast indistinguishable (0.907/0.912, 0.853/0.853, 0.918/0.918). Timing 6.43->4.16s on the T2=40 nudged phase (contended GPU, relative). --holofast flag, default off; queued ablation arms deliberately stay on orig for fidelity. Co-Authored-By: Claude Fable 5 Claude-Session: https://claude.ai/code/session_014FAPDWQ49M5Ye3NpTndTpn --- ep_run/holo_fast_parity.log | 6 ++++++ 1 file changed, 6 insertions(+) create mode 100644 ep_run/holo_fast_parity.log (limited to 'ep_run/holo_fast_parity.log') diff --git a/ep_run/holo_fast_parity.log b/ep_run/holo_fast_parity.log new file mode 100644 index 0000000..1aedf66 --- /dev/null +++ b/ep_run/holo_fast_parity.log @@ -0,0 +1,6 @@ +/home/yurenh2/miniconda3/lib/python3.13/site-packages/torch/autograd/graph.py:865: UserWarning: Attempting to run cuBLAS, but there was no current CUDA context! Attempting to set the primary context... (Triggered internally at /pytorch/aten/src/ATen/cuda/CublasHandlePool.cpp:330.) + return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass +parity: rel_diff=4.48e-01 cos=0.89945102 t_best 20 vs 20 +orig: 6.433s (best of 3) +fast: 4.163s (best of 3) +HOLO_FAST_PARITY_DONE -- cgit v1.2.3