CUDA_VISIBLE_DEVICES=3,4,6 bash /mnt/volumes/base-ov-ali-sh-mix/yjl3/python_project/lmeval_project/eval_script/lmeval_to_eval.sh \ --model_path /lpai/inputs/models/Qwen__Qwen3_5-9B/26-07-18-1215 \ --model_name Qwen__Qwen3_5-9B-think-math_t_n_p \ --python /mnt/volumes/base-ov-ali-sh-mix/yjl3/miniforge3/envs/lmeval_vllm13/bin/python \ --tasks hendrycks_math \ --data_parallel_size 3 \ --apply_chat_template 1 \ --enable_thinking 0 \ --system_instruction "You are a helpful assistant." \ --think_end_token "</think>" \ --batch_size 16 \ --model_be vllm \ --gen_kwargs "do_sample=True,temperature=0.6,top_p=0.95,top_k=20,min_p=0.0,presence_penalty=0.0,repetition_penalty=1.0,max_gen_toks=16384" \ --gpu_memory_utilization 0.85 \ --log_samples 1

CUDA_VISIBLE_DEVICES=3,4,6 bash /mnt/volumes/base-ov-ali-sh-mix/yjl3/python_project/lmeval_project/eval_script/lmeval_to_eval.sh \ --model_path /lpai/inputs/models/Qwen__Qwen3_5-9B/26-07-18-1215 \ --model_name Qwen__Qwen3_5-9B-think-math_t_n_p \ --python /mnt/volumes/base-ov-ali-sh-mix/yjl3/miniforge3/envs/lmeval_vllm13/bin/python \ --tasks hendrycks_math \ --data_parallel_size 3 \ --apply_chat_template 1 \ --enable_thinking 0 \ --system_instruction "You are a helpful assistant." \ --think_end_token "</think>" \ --batch_size 16 \ --model_be vllm \ --gen_kwargs "do_sample=True,temperature=0.6,top_p=0.95,top_k=20,min_p=0.0,presence_penalty=0.0,repetition_penalty=1.0,max_gen_toks=16384" \ --gpu_memory_utilization 0.85 \ --log_samples 1