From 66e0d8b9fd4d0f7a2231d689c055e26fdf1cf04a Mon Sep 17 00:00:00 2001 From: YurenHao0426 Date: Sat, 13 Jun 2026 12:35:36 -0500 Subject: rrm workspace: TRM/HRM/SRM code, Maze dataset, dynamical-analysis pipeline Curated export for clone-and-run Maze training (2x A6000) + diagnostics. trm/hrm pretrain.py carry trajectory-augmentation code (backward-compatible). Heavy artifacts (checkpoints/wandb/npz) gitignored; see PROVENANCE.md. Co-Authored-By: Claude Fable 5 --- scripts/run_trm_sudoku.sh | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+) create mode 100755 scripts/run_trm_sudoku.sh (limited to 'scripts/run_trm_sudoku.sh') diff --git a/scripts/run_trm_sudoku.sh b/scripts/run_trm_sudoku.sh new file mode 100755 index 0000000..ae6db21 --- /dev/null +++ b/scripts/run_trm_sudoku.sh @@ -0,0 +1,21 @@ +#!/usr/bin/env bash +# 启动 TRM Sudoku 1k 训练 (TRM 官方 pretrain_mlp_t_sudoku 配置) +# 单 GPU L40S 48GB 约 18h; A6000 应该接近 +set -euo pipefail +REPO_ROOT="$(cd "$(dirname "$0")/.." && pwd)" +source "$(conda info --base)/etc/profile.d/conda.sh" +conda activate rrm + +cd "$REPO_ROOT/trm" +OMP_NUM_THREADS=${OMP_NUM_THREADS:-8} \ +WANDB_MODE=${WANDB_MODE:-online} \ +python pretrain.py \ + arch=trm \ + data_paths="[$REPO_ROOT/data/sudoku-extreme-1k-aug-1000]" \ + evaluators="[]" \ + epochs=50000 eval_interval=5000 \ + lr=1e-4 puzzle_emb_lr=1e-4 weight_decay=1.0 puzzle_emb_weight_decay=1.0 \ + arch.mlp_t=True arch.pos_encodings=none \ + arch.L_layers=2 \ + arch.H_cycles=3 arch.L_cycles=6 \ + +run_name=pretrain_mlp_t_sudoku ema=True "$@" -- cgit v1.2.3