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 --- research/flossing/run_trm_cf_ef.sh | 39 ++++++++++++++++++++++++++++++++++++++ 1 file changed, 39 insertions(+) create mode 100755 research/flossing/run_trm_cf_ef.sh (limited to 'research/flossing/run_trm_cf_ef.sh') diff --git a/research/flossing/run_trm_cf_ef.sh b/research/flossing/run_trm_cf_ef.sh new file mode 100755 index 0000000..222a300 --- /dev/null +++ b/research/flossing/run_trm_cf_ef.sh @@ -0,0 +1,39 @@ +#!/bin/bash +# E: α=3 λ*=0 (calculated TRM-equivalent of HRM α=10) +# F: α=2 λ*=0 (slightly weaker) +set -e +cd /home/yurenh2/rrm/research/flossing +source /home/yurenh2/miniconda3/etc/profile.d/conda.sh +conda activate rrm +export CUDA_VISIBLE_DEVICES=0 + +PREV_RUNNER_PID=$1 +echo "[$(date '+%H:%M:%S')] EF-runner waiting for prev runner PID $PREV_RUNNER_PID..." >> step5_runner.log +while kill -0 "$PREV_RUNNER_PID" 2>/dev/null; do + sleep 30 +done +echo "[$(date '+%H:%M:%S')] Previous runner done. Starting E/F (α=3/2 calculated equivalents)" >> step5_runner.log + +CKPT_ROOT="/home/yurenh2/rrm/trm/checkpoints/Sudoku-extreme-1k-aug-1000-ACT-torch/pretrain_mlp_t_sudoku_singleGPU" +SRC_CKPT="step_26041" + +run_cf() { + local TAG=$1; local ALPHA=$2; local LSTAR=$3 + local OUT=step5_${TAG}_trm_${SRC_CKPT}_alpha${ALPHA}_lstar${LSTAR}.json + local LOG=step5_${TAG}.log + if [ -f "$OUT" ]; then echo "skip $TAG"; return; fi + echo "[$(date '+%H:%M:%S')] Starting $TAG (α=$ALPHA λ*=$LSTAR) from $SRC_CKPT" >> step5_runner.log + python step5_train_trm_cf.py \ + --ckpt-root "$CKPT_ROOT" --ckpt-name "$SRC_CKPT" \ + --n-steps 500 --batch-size 8 \ + --alpha-rf $ALPHA --rf-mode fixed --lambda-star $LSTAR \ + --k-lyap 2 --lyap-act-steps 4 \ + --eval-every 100 --eval-n 512 --eval-batch-size 32 \ + --out $OUT > $LOG 2>&1 + echo "[$(date '+%H:%M:%S')] $TAG done" >> step5_runner.log +} + +run_cf E 3 0 +run_cf F 2 0 + +echo "[$(date '+%H:%M:%S')] EF complete" >> step5_runner.log -- cgit v1.2.3