rl4phyx-backup / root_scripts /resume_training.sh
YUNTA88's picture
Upload root_scripts/resume_training.sh with huggingface_hub
215f04b verified
#!/bin/bash
cd /workspace/rl4phyx/RL4Phyx/SFT
export CUDA_VISIBLE_DEVICES=0,1,3,4
export PYTHONUNBUFFERED=1
TIMESTAMP=$(date +%Y%m%d_%H%M%S)
LOG="/workspace/rl4phyx/logs/train_lora_phyx_math_f_resume_${TIMESTAMP}.log"
echo "Resuming lora_phyx_math_f from checkpoint-1460 at $(date)" > "$LOG"
torchrun --nproc_per_node=4 --master_port=29500 train_sft_phyx_math_lora_freeze.py >> "$LOG" 2>&1
echo "TRAINING_COMPLETE at $(date)" >> "$LOG"