add scripts

scripts/atom_bond_137k.sh

@@ -0,0 +1,54 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_atom_bond_137k
+train_path=../data/atom_bond_137k/train.csv
+val_path=../data/atom_bond_137k/val.csv
+test_path=../data/atom_bond_137k/test.csv
+train_constraints_path=../data/atom_bond_137k/train_constraints.csv
+val_constraints_path=../data/atom_bond_137k/val_constraints.csv
+test_constraints_path=../data/atom_bond_137k/test_constraints.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--constraints_path $train_constraints_path \
+--separate_val_constraints_path $val_constraints_path \
+--separate_test_constraints_path $val_constraints_path \
+--num_iters 30 \
+--epochs 50 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers  hidden_size ffn_hidden_size dropout \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir \
+--adding_h \
+--is_atom_bond_targets \
+--no_shared_atom_bond_ffn \
+--no_adding_bond_types
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--constraints_path $train_constraints_path \
+--separate_val_constraints_path $val_constraints_path \
+--separate_test_constraints_path $test_constraints_path \
+--epochs 50 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--adding_h \
+--is_atom_bond_targets \
+--no_shared_atom_bond_ffn \
+--no_adding_bond_types \
+--ensemble_size 5 \
+--save_preds \
+--extra_metrics mae \
+--show_individual_scores

scripts/barriers_cycloadd.sh

@@ -0,0 +1,45 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_barriers_cycloadd
+train_path=../data/barriers_cycloadd/train.csv
+val_path=../data/barriers_cycloadd/val.csv
+test_path=../data/barriers_cycloadd/test.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--num_iters 100 \
+--epochs 200 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers hidden_size ffn_hidden_size dropout max_lr final_lr init_lr batch_size warmup_epochs \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir \
+--reaction \
+--explicit_h \
+--smiles_column rxn_smiles \
+--target_columns G_act
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--epochs 200 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--reaction \
+--explicit_h \
+--ensemble_size 5 \
+--save_preds \
+--extra_metrics mae \
+--smiles_column rxn_smiles \
+--target_columns G_act
+

scripts/barriers_e2.sh

@@ -0,0 +1,41 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_barriers_e2
+train_path=../data/barriers_e2/train.csv
+val_path=../data/barriers_e2/val.csv
+test_path=../data/barriers_e2/test.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--num_iters 100 \
+--epochs 200 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers hidden_size ffn_hidden_size dropout max_lr final_lr init_lr batch_size warmup_epochs \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir \
+--reaction \
+--explicit_h 
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--epochs 200 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--reaction \
+--explicit_h \
+--ensemble_size 5 \
+--save_preds \
+--extra_metrics mae
+

scripts/barriers_rdb7.sh

@@ -0,0 +1,41 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_barriers_rdb7
+train_path=../data/barriers_rdb7/train.csv
+val_path=../data/barriers_rdb7/val.csv
+test_path=../data/barriers_rdb7/test.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--num_iters 30 \
+--epochs 50 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers  hidden_size ffn_hidden_size dropout \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir \
+--reaction \
+--explicit_h 
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--epochs 50 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--reaction \
+--explicit_h \
+--ensemble_size 5 \
+--save_preds \
+--extra_metrics mae 
+

scripts/barriers_rgd1.sh

@@ -0,0 +1,41 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_barriers_rgd1
+train_path=../data/barriers_rgd1/train.csv
+val_path=../data/barriers_rgd1/val.csv
+test_path=../data/barriers_rgd1/test.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--num_iters 30 \
+--epochs 50 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers  hidden_size ffn_hidden_size dropout \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir \
+--reaction \
+--explicit_h 
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--epochs 50 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--reaction \
+--explicit_h \
+--ensemble_size 5 \
+--save_preds \
+--extra_metrics mae 
+

scripts/barriers_sn2.sh

@@ -0,0 +1,41 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_barriers_sn2
+train_path=../data/barriers_sn2/train.csv
+val_path=../data/barriers_sn2/val.csv
+test_path=../data/barriers_sn2/test.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--num_iters 100 \
+--epochs 200 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers hidden_size ffn_hidden_size dropout max_lr final_lr init_lr batch_size warmup_epochs \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir \
+--reaction \
+--explicit_h 
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--epochs 200 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--reaction \
+--explicit_h \
+--ensemble_size 5 \
+--save_preds \
+--extra_metrics mae
+

scripts/bde.sh

@@ -0,0 +1,45 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_bde
+train_path=../data/bde/train.csv
+val_path=../data/bde/val.csv
+test_path=../data/bde/test.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--num_iters 30 \
+--epochs 50 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers  hidden_size ffn_hidden_size dropout \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir \
+--adding_h \
+--is_atom_bond_targets \
+--no_shared_atom_bond_ffn \
+--no_adding_bond_types
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--epochs 50 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--adding_h \
+--is_atom_bond_targets \
+--no_shared_atom_bond_ffn \
+--no_adding_bond_types \
+--ensemble_size 5 \
+--save_preds \
+--extra_metrics mae
+

scripts/bde_charges.sh

@@ -0,0 +1,54 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_bde_charges
+train_path=../data/bde_charges/train.csv
+val_path=../data/bde_charges/val.csv
+test_path=../data/bde_charges/test.csv
+train_constraints_path=../data/bde_charges/train_constraints.csv
+val_constraints_path=../data/bde_charges/val_constraints.csv
+test_constraints_path=../data/bde_charges/test_constraints.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--constraints_path $train_constraints_path \
+--separate_val_constraints_path $val_constraints_path \
+--separate_test_constraints_path $val_constraints_path \
+--num_iters 30 \
+--epochs 50 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers  hidden_size ffn_hidden_size dropout \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir \
+--adding_h \
+--is_atom_bond_targets \
+--no_shared_atom_bond_ffn \
+--no_adding_bond_types
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--constraints_path $train_constraints_path \
+--separate_val_constraints_path $val_constraints_path \
+--separate_test_constraints_path $test_constraints_path \
+--epochs 50 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--adding_h \
+--is_atom_bond_targets \
+--no_shared_atom_bond_ffn \
+--no_adding_bond_types \
+--ensemble_size 5 \
+--save_preds \
+--extra_metrics mae \
+--show_individual_scores

scripts/charges_eps_4.sh

@@ -0,0 +1,63 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_charges_eps_4
+train_path=../data/charges_eps_4/train.csv
+val_path=../data/charges_eps_4/val.csv
+test_path=../data/charges_eps_4/test.csv
+train_constraints_path=../data/charges_eps_4/train_constraints.csv
+val_constraints_path=../data/charges_eps_4/val_constraints.csv
+test_constraints_path=../data/charges_eps_4/test_constraints.csv
+
+external_test_path=../data/charges_eps_4/external_test_set.csv
+external_test_constraints_path=../data/charges_eps_4/external_test_set_constraints.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--constraints_path $train_constraints_path \
+--separate_val_constraints_path $val_constraints_path \
+--separate_test_constraints_path $val_constraints_path \
+--num_iters 30 \
+--epochs 50 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers  hidden_size ffn_hidden_size dropout \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir \
+--adding_h \
+--is_atom_bond_targets \
+--no_shared_atom_bond_ffn \
+--no_adding_bond_types
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--constraints_path $train_constraints_path \
+--separate_val_constraints_path $val_constraints_path \
+--separate_test_constraints_path $test_constraints_path \
+--epochs 50 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--adding_h \
+--is_atom_bond_targets \
+--no_shared_atom_bond_ffn \
+--no_adding_bond_types \
+--ensemble_size 5 \
+--save_preds \
+--extra_metrics mae
+
+#Predict on external test set
+python $chemprop_dir/predict.py \
+--test_path $external_test_path \
+--constraints_path $external_test_constraints_path \
+--preds_path $results_dir/preds_external_test.csv \
+--checkpoint_dir $results_dir

scripts/charges_eps_78.sh

@@ -0,0 +1,63 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_charges_eps_78
+train_path=../data/charges_eps_78/train.csv
+val_path=../data/charges_eps_78/val.csv
+test_path=../data/charges_eps_78/test.csv
+train_constraints_path=../data/charges_eps_78/train_constraints.csv
+val_constraints_path=../data/charges_eps_78/val_constraints.csv
+test_constraints_path=../data/charges_eps_78/test_constraints.csv
+
+external_test_path=../data/charges_eps_78/external_test_set.csv
+external_test_constraints_path=../data/charges_eps_78/external_test_set_constraints.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--constraints_path $train_constraints_path \
+--separate_val_constraints_path $val_constraints_path \
+--separate_test_constraints_path $val_constraints_path \
+--num_iters 30 \
+--epochs 50 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers  hidden_size ffn_hidden_size dropout \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir \
+--adding_h \
+--is_atom_bond_targets \
+--no_shared_atom_bond_ffn \
+--no_adding_bond_types
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--constraints_path $train_constraints_path \
+--separate_val_constraints_path $val_constraints_path \
+--separate_test_constraints_path $test_constraints_path \
+--epochs 50 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--adding_h \
+--is_atom_bond_targets \
+--no_shared_atom_bond_ffn \
+--no_adding_bond_types \
+--ensemble_size 5 \
+--save_preds \
+--extra_metrics mae
+
+#Predict on external test set
+python $chemprop_dir/predict.py \
+--test_path $external_test_path \
+--constraints_path $external_test_constraints_path \
+--preds_path $results_dir/preds_external_test.csv \
+--checkpoint_dir $results_dir

scripts/hiv.sh

@@ -0,0 +1,37 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_hiv
+train_path=../data/hiv/train.csv
+val_path=../data/hiv/val.csv
+test_path=../data/hiv/test.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type classification \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--num_iters 30 \
+--epochs 50 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers  hidden_size ffn_hidden_size dropout \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir 
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type classification \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--epochs 50 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--ensemble_size 5 \
+--save_preds \
+--extra_metrics prc-auc
+

scripts/ir.sh

@@ -0,0 +1,35 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_ir
+train_path=../data/ir/train.csv
+val_path=../data/ir/val.csv
+test_path=../data/ir/test.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type spectra \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--num_iters 100 \
+--epochs 200 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers hidden_size ffn_hidden_size dropout max_lr final_lr init_lr batch_size warmup_epochs \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir 
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type spectra \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--epochs 200 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--ensemble_size 5 \
+--save_preds 

scripts/multi_molecule.sh

@@ -0,0 +1,39 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_multi_molecule
+train_path=../data/multi_molecule/train.csv
+val_path=../data/multi_molecule/val.csv
+test_path=../data/multi_molecule/test.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--num_iters 30 \
+--epochs 50 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers  hidden_size ffn_hidden_size dropout \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir \
+--number_of_molecules 2
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--epochs 50 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--number_of_molecules 2 \
+--ensemble_size 5 \
+--save_preds \
+--extra_metrics mae r2
+

scripts/pcba_random.sh

@@ -0,0 +1,37 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_pcba_random
+train_path=../data/pcba_random/train.csv
+val_path=../data/pcba_random/val.csv
+test_path=../data/pcba_random/test.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type classification \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--num_iters 30 \
+--epochs 50 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers  hidden_size ffn_hidden_size dropout \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir 
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type classification \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--epochs 50 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--ensemble_size 5 \
+--save_preds \
+--extra_metrics prc-auc
+

scripts/pcba_random_nans.sh

@@ -0,0 +1,37 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_pcba_random_nans
+train_path=../data/pcba_random_nans/train.csv
+val_path=../data/pcba_random_nans/val.csv
+test_path=../data/pcba_random_nans/test.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type classification \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--num_iters 30 \
+--epochs 50 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers  hidden_size ffn_hidden_size dropout \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir 
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type classification \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--epochs 50 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--ensemble_size 5 \
+--save_preds \
+--extra_metrics prc-auc
+

scripts/pcba_scaffold.sh

@@ -0,0 +1,37 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_pcba_scaffold
+train_path=../data/pcba_scaffold/train.csv
+val_path=../data/pcba_scaffold/val.csv
+test_path=../data/pcba_scaffold/test.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type classification \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--num_iters 30 \
+--epochs 50 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers  hidden_size ffn_hidden_size dropout \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir 
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type classification \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--epochs 50 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--ensemble_size 5 \
+--save_preds \
+--extra_metrics prc-auc
+

scripts/pcqm4mv2.sh

@@ -0,0 +1,36 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_pcqm4mv2
+train_path=../data/pcqm4mv2/train.csv
+val_path=../data/pcqm4mv2/val.csv
+test_path=../data/pcqm4mv2/test.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--num_iters 30 \
+--epochs 50 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers  hidden_size ffn_hidden_size dropout \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir 
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--epochs 50 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--ensemble_size 5 \
+--save_preds \
+--extra_metrics mae 

scripts/qm9_gap.sh

@@ -0,0 +1,39 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_qm9_gap
+train_path=../data/qm9/train.csv
+val_path=../data/qm9/val.csv
+test_path=../data/qm9/test.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--num_iters 30 \
+--epochs 50 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers  hidden_size ffn_hidden_size dropout \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir \
+--target_columns gap
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--epochs 50 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--ensemble_size 5 \
+--save_preds \
+--extra_metrics mae \
+--show_individual_scores \
+--target_columns gap

scripts/qm9_multitask.sh

@@ -0,0 +1,37 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_qm9_multitask
+train_path=../data/qm9/train.csv
+val_path=../data/qm9/val.csv
+test_path=../data/qm9/test.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--num_iters 30 \
+--epochs 50 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers  hidden_size ffn_hidden_size dropout \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir 
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--epochs 50 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--ensemble_size 5 \
+--save_preds \
+--extra_metrics mae \
+--show_individual_scores

scripts/qm9_u0.sh

@@ -0,0 +1,39 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_qm9_u0
+train_path=../data/qm9/train.csv
+val_path=../data/qm9/val.csv
+test_path=../data/qm9/test.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--num_iters 30 \
+--epochs 50 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers  hidden_size ffn_hidden_size dropout \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir \
+--target_columns u0_atom 
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--epochs 50 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--ensemble_size 5 \
+--save_preds \
+--extra_metrics mae \
+--show_individual_scores \
+--target_columns u0_atom 

scripts/sampl.sh

@@ -0,0 +1,83 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_sampl
+results_dir2=results_sampl_production
+train_path=../data/logP/train.csv
+val_path=../data/logP/val.csv
+test_path=../data/logP/test.csv
+path=../data/logP/logP_without_overlap.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--num_iters 30 \
+--epochs 50 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers  hidden_size ffn_hidden_size dropout \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir 
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--epochs 50 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--ensemble_size 5 \
+--save_preds \
+--extra_metrics mae
+
+#Train production model
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $path \
+--separate_val_path $path \
+--separate_test_path $path \
+--epochs 40 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir2 \
+--ensemble_size 5 
+
+#Predict on Sample 6
+python $chemprop_dir/predict.py \
+--test_path "../data/logP/sampl6_experimental.csv" \
+--preds_path $results_dir2/pred_SAMPL6.csv \
+--checkpoint_dir $results_dir2 \
+--smiles_column "Isomeric SMILES"
+
+echo SAMPL6  >> $results_dir2/sampl.csv
+python -c 'import pandas as pd; from sklearn import metrics; print("rmse", metrics.mean_squared_error(pd.read_csv("results_sampl_production/pred_SAMPL6.csv")["logP"],pd.read_csv("../data/logP/sampl6_experimental.csv")["logP mean"],squared=False))' >> $results_dir2/sampl.csv
+
+#Predict on Sample 7
+python $chemprop_dir/predict.py \
+--test_path "../data/logP/sampl7_experimental.csv" \
+--preds_path $results_dir2/pred_SAMPL7.csv \
+--checkpoint_dir $results_dir2 \
+--smiles_column "Isomeric SMILES"
+
+echo SAMPL7 >> $results_dir2/sampl.csv
+python -c 'import pandas as pd; from sklearn import metrics; print("rmse", metrics.mean_squared_error(pd.read_csv("results_sampl_production/pred_SAMPL7.csv")["logP"],pd.read_csv("../data/logP/sampl7_experimental.csv")["logP mean"],squared=False))' >> $results_dir2/sampl.csv
+
+#Predict on Sample 9
+python $chemprop_dir/predict.py \
+--test_path "../data/logP/sampl9_experimental.csv" \
+--preds_path $results_dir2/pred_SAMPL9.csv \
+--checkpoint_dir $results_dir2 \
+--smiles_column smiles
+
+echo SAMPL9 >> $results_dir2/sampl.csv
+python -c 'import pandas as pd; from sklearn import metrics; print("rmse", metrics.mean_squared_error(pd.read_csv("results_sampl_production/pred_SAMPL9.csv")["logP"],pd.read_csv("../data/logP/sampl9_experimental.csv")["new_logPexp_reviewed"],squared=False))' >> $results_dir2/sampl.csv
+
+echo "Saved results to" $results_dir2"/sampl.csv"
+cat  >> $results_dir2/sampl.csv

scripts/timing.sh

@@ -0,0 +1,58 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+save_dir=results_timing
+data_dir=../data/timing
+
+#100k
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $data_dir/qm9_100k.csv \
+--save_dir $save_dir/qm9_100k \
+--aggregation norm \
+--depth 4 \
+--ffn_num_layers 2 \
+--hidden_size 1000 \
+--ffn_hidden_size 1000 \
+--epochs 50 
+
+python $chemprop_dir/predict.py \
+--test_path $data_dir/qm9_100k.csv \
+--preds_path $save_dir/qm9_100k/preds/preds.csv \
+--checkpoint_dir $save_dir/qm9_100k 
+
+
+#10k
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $data_dir/qm9_10k.csv \
+--save_dir $save_dir/qm9_10k \
+--aggregation norm \
+--depth 4 \
+--ffn_num_layers 2 \
+--hidden_size 1000 \
+--ffn_hidden_size 1000 \
+--epochs 50
+
+python $chemprop_dir/predict.py \
+--test_path $data_dir/qm9_10k.csv \
+--preds_path $save_dir/qm9_10k/preds/preds.csv \
+ --checkpoint_dir $save_dir/qm9_10k 
+
+#1k
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $data_dir/qm9_1k.csv \
+--save_dir $save_dir/qm9_1k \
+--aggregation norm \
+--depth 4 \
+--ffn_num_layers 2 \
+--hidden_size 1000 \
+--ffn_hidden_size 1000 \
+--epochs 50 
+
+python $chemprop_dir/predict.py \
+--test_path $data_dir/qm9_1k.csv \
+--preds_path $save_dir/qm9_1k/preds/preds.csv \
+--checkpoint_dir $save_dir/qm9_1k 

scripts/uncertainty_ensemble.sh

@@ -0,0 +1,49 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_uncertainty_ensemble
+train_path=../data/uncertainty/train.csv
+val_path=../data/uncertainty/val.csv
+test_path=../data/uncertainty/test.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--num_iters 30 \
+--epochs 50 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers  hidden_size ffn_hidden_size dropout \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--epochs 50 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--ensemble_size 5 
+
+#Predict, analyze uncertainty
+python $chemprop_dir/predict.py \
+--test_path $test_path \
+--preds_path $results_dir/test_preds_unc_ensemble.csv \
+--checkpoint_dir $results_dir \
+--uncertainty_method ensemble \
+--calibration_method zscaling \
+--calibration_path $val_path \
+--regression_calibrator_metric stdev \
+--calibration_interval_percentile 95 \
+--evaluation_methods nll spearman ence miscalibration_area \
+--evaluation_scores_path $results_dir/unc_eval_scores_ensemble.csv		
+
+cat $results_dir/unc_eval_scores_ensemble.csv

scripts/uncertainty_evidential.sh

@@ -0,0 +1,50 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_uncertainty_evidential
+train_path=../data/uncertainty/train.csv
+val_path=../data/uncertainty/val.csv
+test_path=../data/uncertainty/test.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--num_iters 30 \
+--epochs 50 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers  hidden_size ffn_hidden_size dropout \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir \
+--loss_function evidential
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--epochs 50 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--loss_function evidential
+
+#Predict, analyze uncertainty
+python $chemprop_dir/predict.py \
+--test_path $test_path \
+--preds_path $results_dir/test_preds_unc_evidential.csv \
+--checkpoint_dir $results_dir \
+--uncertainty_method evidential_total \
+--calibration_method zscaling \
+--calibration_path $val_path \
+--regression_calibrator_metric stdev \
+--calibration_interval_percentile 95 \
+--evaluation_methods nll spearman ence miscalibration_area \
+--evaluation_scores_path $results_dir/unc_eval_scores_evidential.csv		
+
+cat $results_dir/unc_eval_scores_evidential.csv

scripts/uncertainty_mve.sh

@@ -0,0 +1,50 @@
+#!/bin/bash
+
+chemprop_dir=../../../chemprop  # location of chemprop directory, CHANGE ME
+
+results_dir=results_uncertainty_mve
+train_path=../data/uncertainty/train.csv
+val_path=../data/uncertainty/val.csv
+test_path=../data/uncertainty/test.csv
+
+#Hyperparameter optimization
+python $chemprop_dir/hyperparameter_optimization.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $val_path \
+--num_iters 30 \
+--epochs 50 \
+--aggregation norm \
+--search_parameter_keywords depth ffn_num_layers  hidden_size ffn_hidden_size dropout \
+--config_save_path $results_dir/config.json \
+--hyperopt_checkpoint_dir $results_dir \
+--log_dir $results_dir \
+--loss_function mve
+
+#Training with optimized hyperparameters
+python $chemprop_dir/train.py \
+--dataset_type regression \
+--data_path $train_path \
+--separate_val_path $val_path \
+--separate_test_path $test_path \
+--epochs 50 \
+--aggregation norm \
+--config_path $results_dir/config.json \
+--save_dir $results_dir \
+--loss_function mve
+
+#Predict, analyze uncertainty
+python $chemprop_dir/predict.py \
+--test_path $test_path \
+--preds_path $results_dir/test_preds_unc_mve.csv \
+--checkpoint_dir $results_dir \
+--uncertainty_method mve \
+--calibration_method zscaling \
+--calibration_path $val_path \
+--regression_calibrator_metric stdev \
+--calibration_interval_percentile 95 \
+--evaluation_methods nll spearman ence miscalibration_area \
+--evaluation_scores_path $results_dir/unc_eval_scores_mve.csv		
+
+cat $results_dir/unc_eval_scores_mve.csv