#g 复现thinkprune才会使用到该代码,其余的使用thinkprune数据训练的模型,不需要设置remaining=thinkrpune # Qwen2.5-Math-Instruct Series PROMPT_TYPE="deepseek3" # PROMPT_TYPE="qwen25-math-cot" # Qwen2.5-Math-1.5B-Instruct DATA_NAME="math500" # export CUDA_VISIBLE_DEVICES="4,5,6,7" export CUDA_VISIBLE_DEVICES="0,1,2,3" # export CUDA_VISIBLE_DEVICES="0,1,2,3,4,5,6,7" export remaining=default bash delete_file.sh ./start_positions.pt bash delete_file.sh ./early_positions.pt MODEL_NAME_OR_PATH='/mnt/lyc/wuxinrui/ThinkPrune/models/DeepSeek-R1-Distill-Qwen-1.5B-thinkprune-iter2k' PARENT_DIR=$(dirname "$MODEL_NAME_OR_PATH") # 获取父目录 MODEL_NAME=$(basename "$PARENT_DIR") # 获取父目录的最后一部分 echo MODEL_NAME: $MODEL_NAME export PE_MODE=default export position=ori export tip=thinkprune export stage=2 export mode=TIP-$tip-STAGE-$stage export model=MODEL-$MODEL_NAME export modelname=MODEL-$MODEL_NAME-TIP-$tip-STAGE-$stage-DATA-$DATA_NAME bash ./sh/remaining.sh $PROMPT_TYPE $MODEL_NAME_OR_PATH $DATA_NAME