| MODEL_NAME="Qwen/Qwen2.5-7B-Instruct" | |
| MAX_COMPLETION_TOKENS=2048 | |
| METHOD="Majority" | |
| DATASET="mathhard_train" | |
| BUDGET=8 | |
| # OUTPUT_PATH="/home/tianqiu/tts_schedule/batch_infer/results/test/output" | |
| export CUDA_VISIBLE_DEVICES=4,5,6,7 | |
| python /home/tianqiu/tts_schedule/batch_infer/src/run_pipeline.py --model_name $MODEL_NAME --max_completion_tokens $MAX_COMPLETION_TOKENS --method $METHOD --dataset $DATASET --tensor_parallel_size 4 --budget $BUDGET |