lm-evaluation-harness/run_scripts/eval_base.sh

#!/bin/bash 
export HTTP_PROXY=http://sys-proxy-rd-relay.byted.org:8118
export http_proxy=http://sys-proxy-rd-relay.byted.org:8118
export https_proxy=http://sys-proxy-rd-relay.byted.org:8118
export no_proxy="$no_proxy,.byteintl.net"
export HF_ENDPOINT=https://hf-mirror.com

model=/mnt/bn/life-mllm/users/cxr/quantization/models/Qwen2.5-7B
# output_dir=$3
task=hellaswag
# task=hellaswag,rte,mmlu
lm_eval --model hf \
    --model_args pretrained=${model} \
    --tasks hellaswag \
    --device cuda:0 \
    --batch_size 64 \
    --num_fewshot 10 \
    --output_path results2/base/base_${model}.json

# lm_eval --model hf \
#     --model_args pretrained=${model} \
#     --tasks arc_challenge \
#     --device cuda:1 \
#     --batch_size 16 \
#     --num_fewshot 25 \
#     --output_path results2/base/base_${model}.json