run-selection-strategy-nq-dev-seal-2.sh

#!/bin/bash

export gpu_=$CUDA_VISIBLE_DEVICES

#strict_positive_naive_damaging_remove_damage_relevant.json
#strict_positive_strict_damaging_remove_damage_relevant.json


for file in /data/philhoon-relevance/FiD/open_domain_data/NQ_DEV_SEAL_SELECTION/*.json
do

    if [[ $file == *"strict_positive_"* ]] && [[ $file == *"damaging_remove_damage_relevant.json" ]]; then
        name=${file##*/}
        base=${name%.json}

        echo "CUDA_VISIBLE_DEVICES="$gpu_" python test_reader-slurm.py \
            --model_path /data/philhoon-relevance/FiD/pretrained_models/nq_reader_large \
            --eval_data "$file" \
            --write_results \
            --per_gpu_batch_size 9 \
            --n_context 100 \
            --name "$base" \
            --checkpoint_dir /data/philhoon-relevance/FiD/results/NQ_DEV_SEAL_SELECTION/
        "
        CUDA_VISIBLE_DEVICES="$gpu_" python test_reader-slurm.py \
            --model_path /data/philhoon-relevance/FiD/pretrained_models/nq_reader_large \
            --eval_data "$file" \
            --write_results \
            --per_gpu_batch_size 9 \
            --n_context 100 \
            --name "$base" \
            --checkpoint_dir /data/philhoon-relevance/FiD/results/NQ_DEV_SEAL_SELECTION/
    fi
done

# Script for selection strategies
# inputs : all json files on /data/philhoon-relevance/FiD/open_domain_data/NQ_KILT_BM25_SELECTION
# outputs : /data/philhoon-relevance/FiD/results/NQ_KILT_BM25_SELECTION

#CUDA_VISIBLE_DEVICES=1 python test_reader.py \
#        --model_path /data/philhoon-relevance/FiD/pretrained_models/nq_reader_large \
#        --eval_data /data/philhoon-relevance/FiD/open_domain_data/NQ_KILT_BM25_SELECTION/"$dataset" \
#        --write_results \
#        --per_gpu_batch_size 128 \
#        --n_context 5 \
#        --name "$attempt" \
#        --checkpoint_dir /data/philhoon-relevance/FiD/results/NQ_KILT_BM25_SELECTION