arc-c.sh

export HF_ENDPOINT=https://hf-mirror.com

# args='{"pretrained":"/path/to/LLaMA-Factory-0.5.3/our_trained_models/7.17.llama2-7b-newarch_2layer256_relu-applyrope-huber1-ce0.75_mse0.25-zero2-LA5,6,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31-rank1024-3e4-4wupdate-bsz128-merge_v2_head_7b-nodetach-DS10/checkpoint-15000","special_training_attn_layers":[5,6,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],"hidden_size_low_rank":1024}'

args='{"pretrained":"/path/to/original_weight/llama-2-7b-hf"}'

CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 accelerate launch --main_process_port 29508 -m lm_eval --model hf \
    --model_args $args \
    --tasks arc_challenge \
    --batch_size 16 \
    --num_fewshot 25