-
Notifications
You must be signed in to change notification settings - Fork 3
/
train_k_fold.sh
19 lines (14 loc) · 3.17 KB
/
train_k_fold.sh
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
array=( 0 1 2 3 4 5 6 7 8 9 )
data_dir=/data
for i in "${array[@]}"
do
python train.py --data_path ${data_dir}/mapped_Buchwald_Hartwig/train_cv_${i}.csv --features_path ${data_dir}/mapped_Buchwald_Hartwig/train_cv_${i}_feat_rdkit.csv --dataset_type regression --separate_val_path ${data_dir}/mapped_Buchwald_Hartwig/test_cv_${i}.csv --separate_test_path ${data_dir}/mapped_Buchwald_Hartwig/test_cv_${i}.csv --separate_val_features_path ${data_dir}/mapped_Buchwald_Hartwig/test_cv_${i}_feat_rdkit.csv --separate_test_features_path ${data_dir}/mapped_Buchwald_Hartwig/test_cv_${i}_feat_rdkit.csv --save_dir ./output/Buchwald_Hartwig_f${i}/ --reaction --epochs 110 --metric r2 --batch_size 64 --hidden_size 1000 --bias --depth 3 --depth_diff 1 --ffn_num_layers 3 --dropout 0.25 --activation LeakyReLU --no_features_scaling
python predict.py --test_path ${data_dir}/mapped_Buchwald_Hartwig/train_cv_${i}.csv --features_path ${data_dir}/mapped_Buchwald_Hartwig/train_cv_${i}_feat_rdkit.csv --checkpoint_dir ./output/Buchwald_Hartwig_f${i}/fold_0/model_0/ --preds_path ./output/Buchwald_Hartwig_f${i}/train_preds.csv
python predict.py --test_path ${data_dir}/mapped_Buchwald_Hartwig/test_cv_${i}.csv --features_path ${data_dir}/mapped_Buchwald_Hartwig/test_cv_${i}_feat_rdkit.csv --checkpoint_dir ./output/Buchwald_Hartwig_f${i}/fold_0/model_0/ --preds_path ./D-MPNN/output/Buchwald_Hartwig_f${i}/test_preds.csv --middle_representation_path ./output/Buchwald_Hartwig_f${i}/middle_layer_representations.csv --last_ffn_representation_path ./output/Buchwald_Hartwig_f${i}/last_layer_representations.csv
done
for i in "${array[@]}"
do
python train.py --data_path ${data_dir}/mapped_Suzuki_Miyaura/train_cv_${i}.csv --features_path ${data_dir}/mapped_Suzuki_Miyaura/train_f${i}_feat_rdkit.csv --dataset_type regression --separate_val_path ${data_dir}/mapped_Suzuki_Miyaura/test_cv_${i}.csv --separate_test_path ${data_dir}/mapped_Suzuki_Miyaura/test_cv_${i}.csv --separate_val_features_path ${data_dir}/mapped_Suzuki_Miyaura/test_f${i}_feat_rdkit_reactant.csv --separate_test_features_path ${data_dir}/mapped_Suzuki_Miyaura/test_f${i}_feat_rdkit_reactant.csv --save_dir /home/dzvinka/D-MPNN/output/Suzuki_myiara_f${i}/ --reaction --epochs 65 --metric r2 --batch_size 256 --hidden_size 1000 --diff_hidden_size 2000 --bias --depth 1 --depth_diff 0 --ffn_num_layers 3 --dropout 0.1 --activation LeakyReLU --no_features_scaling
python predict.py --test_path ${data_dir}/mapped_Suzuki_Miyaura/train_cv_${i}.csv --features_path ${data_dir}/mapped_Suzuki_Miyaura/train_cv_${i}_feat_rdkit.csv --checkpoint_dir ./output/Suzuki_Miyaura_f${i}/fold_0/model_0/ --preds_path ./output/Suzuki_Miyaura_f${i}/train_preds.csv
python predict.py --test_path ${data_dir}/mapped_Suzuki_Miyaura/test_cv_${i}.csv --features_path ${data_dir}/mapped_Suzuki_Miyaura/test_cv_${i}_feat_rdkit.csv --checkpoint_dir ./output/Suzuki_Miyaura_f${i}/fold_0/model_0/ --preds_path ./output/Suzuki_Miyaura_f${i}/test_preds.csv --middle_representation_path ./output/Suzuki_Miyaura_f${i}/middle_layer_representations.csv --last_ffn_representation_path ./output/Suzuki_Miyaura_f${i}/last_layer_representations.csv
done