| 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="$no_proxy,.byteintl.net" | |
| export HF_ENDPOINT=https://hf-mirror.com | |
| model_id=/mnt/bn/life-mllm/users/cxr/quantization/models/Qwen/Qwen2.5-7B | |
| model_name=$(basename "$model_id") | |
| model=/mnt/bn/life-mllm/users/cxr/quantization/models/${model_name}-quantization-fg | |
| # output_dir=Alpha_values_mlp | |
| # bit_layers_dir=/mnt/bn/life-mllm/users/cxr/quantization/quantization_metric/bit_layers | |
| # result_dir=/mnt/bn/life-mllm/users/cxr/quantization/lm-evaluation-harness/results | |
| start=$(date +%s.%N) | |
| # rm -rf $model | |
| cd ../quantization_metric | |
| self_attn_layer_to_quant="4 1 2 8 23" | |
| mlp_layer_to_quant="27 16 19 17 25" | |
| python -u main_low.py --cuda_id 1 --bit_layers --save_dir ${model} --model_id $model_id --mlp_layer_to_quant "${mlp_layer_to_quant}" | |
| cd ../lm-evaluation-harness | |
| bash run_scripts/eval_base_fg.sh ${model} "fg" | |
| rm -rf ${model} | |
| end=$(date +%s.%N) | |
| runtime=$(awk "BEGIN {print $end - $start}") | |
| echo "Execution time: $runtime seconds" | |