script/eval_lm.sh

export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7

model_name_or_path="meta-llama/Llama-2-7b-chat-hf"
# model_name_or_path="Yukang/LongAlpaca-7B-16k"
# model_name_or_path="lmsys/longchat-7b-v1.5-32k"
# model_name_or_path="syzymon/long_llama_code_7b_instruct"
# super_tokenizer_name_or_path="/home/baaiks/ninglu/code/PluginTransformer/data/outputs/90k_0104+8/super_tokenizer"
super_tokenizer_name_or_path="/home/baaiks/ninglu/code/PluginTransformer/data/outputs/90k_0104+12/super_tokenizer"
# super_tokenizer_name_or_path="/home/baaiks/ninglu/code/PluginTransformer/data/outputs/90k_0111+8/super_tokenizer"
# output_dir="data/results/lm/test_1"

# mkdir -p ${output_dir}

# python -m main.eval_lm \
#     --model_name_or_path ${model_name_or_path} \
#     --super_tokenizer_name_or_path ${super_tokenizer_name_or_path} \
#     --super_tokenizer_num_hidden_layers 8 \
#     --device_map "auto" \
#     --dataset_list "pg19" \
#     --output_dir ${output_dir} \
#     --max_length 8192 \
#     --target_length 512 \
#     --compression_enable true \
#     --compression_ratio 16 \
#     | tee "${output_dir}/eval_lm.log"

max_length_lst="4096 8192 16384 32768"
compression_ratio_lst="16"

for max_length in ${max_length_lst}; do
    for compression_ratio in ${compression_ratio_lst}; do
        output_dir="data/results/lm/pg19/llama-12-${compression_ratio}-${max_length}"
        mkdir -p ${output_dir}

        python -m main.eval_lm \
            --model_name_or_path ${model_name_or_path} \
            --super_tokenizer_name_or_path ${super_tokenizer_name_or_path} \
            --super_tokenizer_num_hidden_layers 12 \
            --device_map "auto" \
            --dataset_list "pg19" \
            --output_dir ${output_dir} \
            --max_length ${max_length} \
            --target_length 512 \
            --compression_enable true \
            --compression_ratio ${compression_ratio} \
            --down_scale_method "uniform" \
            | tee "${output_dir}/eval_lm.log"
    done
done