run_official.sh

#!/bin/bash
set -e # exit on error
datadir="Dataset"
dset="Laptop"
finetuned="no-ft"          # "ft", "no-ft"
ftptm="roberta-en"      # if fine-tuned, choose from 'bert-en-base-uncased', 'roberta-en', 'roberta-en-large', 'xlmroberta-xlm-roberta-base', 'bert-multi-base-cased', 'xlnet-xlnet-base-cased'
ptm="roberta"
project_root="/home/P76114511/projects/RoBERTaABSA"
state_dict_path="$project_root/Train/save_models/$ftptm-$dset-FT/$ftptm"
echo == run_official.sh ==
# run this script at project root:
# Note: please activate your venv before running. mb activate robertaabsa
# ft_model_path="$project_root/Train/save_models/$ftptm-$dset-FT/$ftptm"
if [[ $finetuned == "ft" ]]
then
    echo == do finetuning ==
    python3 Train/finetune.py --root_fp=$project_root --model_name=$ftptm --n_epochs=1 --data_dir=$datadir --dataset=$dset # need to use their dataset
    python3 Perturbed-Masking/generate_matrix.py --model_path=$ft_model_path --data_dir=$datadir --dataset=$dset
else
    echo == no finetuning ==
    python3 Perturbed-Masking/generate_matrix.py --model_path=$ptm --data_dir=$datadir --dataset=$dset
fi

echo == training ALSC ==
for layer in {0..12}
do
    echo == training w/ $layer layer ==
    python3 Perturbed-Masking/generate_asgcn.py --layer=$layer --is_finetuned=$finetuned --matrix_folder="$ptm/${dset}_${finetuned}/Train" --root_fp=$project_root
    python3 Perturbed-Masking/generate_asgcn.py --layer=$layer --is_finetuned=$finetuned --matrix_folder="$ptm/${dset}_${finetuned}/Test"  --root_fp=$project_root
    python3 ASGCN/train.py --dataset=asgcn2/$ptm/$layer/${dset}_${finetuned} --save \
    --is_finetuned=$finetuned --ptm_name=$ptm --root_fp=$project_root --layers=$layer
done