run.sh

#!/bin/bash

# make sure to change the environment
module load python/3.7
# source the virtual environment

seed_arr=(13 21 42 87 100)
#seed_arr=(42)
seed_len=${#seed_arr[@]}

#k_arr=(1 2 4 8 16 32 64 128 256 512 1024)
k_arr=(1 2 4 8 16)
k_len=${#k_arr[@]}

path=data/k-shot/CoLA/16-

for ((idx_1=0; idx_1<$seed_len; idx_1++))
do
    for ((idx_2=0; idx_2<$k_len; idx_2++))
    do
        task_id=`expr $idx_1 \* $k_len + $idx_2`
        data=$path${seed_arr[$idx_1]}

        srun python -u run_prompt.py --model_name_or_path roberta-large --task_name cola --data_dir $data   --output_dir outputs_large     --shot_num 16    --max_train_steps 1000     --num_warmup_steps 0     --eval_steps 100     --learning_rate 5e-5     --per_device_train_batch_size 4     --per_device_eval_batch_size 4     --top_k ${k_arr[$idx_2]}     --max_seq_len 128 --template *cls**sent_0*_This_is*mask*.*sep+* --seed ${seed_arr[$idx_1]} --exp_id $task_id --dedup AMuLaP
    done
done