#!/bin/bash source ~/anaconda3/etc/profile.d/conda.sh conda activate pytorch_p38 # multistage-physical for input_type in physical hetero geometric; do CUDA_VISIBLE_DEVICES=2 python train.py --accelerator gpu \ --devices 1 \ --max_epochs 500 \ --precision 16 \ --protein_num_layers 3 \ --ligand_num_layers 3 \ --complex_num_layers 3 \ --protein_node_h_dim 200 32 \ --protein_edge_h_dim 64 2 \ --ligand_node_h_dim 200 32 \ --ligand_edge_h_dim 64 2 \ --complex_node_h_dim 200 32 \ --complex_edge_h_dim 64 2 \ --dataset_name PDBBind \ --input_type multistage-$input_type \ --model_name gvp-multistage \ --data_dir /home/ec2-user/SageMaker/efs/data/PIGNet/data/pdbbind_v2019/scoring \ --residual \ --num_workers 8 \ --lr 1e-4 \ --bs 128 \ --early_stopping_patience 10 \ --default_root_dir /home/ec2-user/SageMaker/efs/model_logs/zichen/PDBBind_MSGVP_$input_type done python evaluate_casf2016.py --model_name gvp-multistage \ --input_type multistage-$input_type \ --num_workers 8 \ --data_dir /home/ec2-user/SageMaker/efs/data/PIGNet/data/casf2016 \ --checkpoint_path /home/ec2-user/SageMaker/efs/model_logs/zichen/PDBBind_MSGVP_$input_type/lightning_logs/version_2 CUDA_VISIBLE_DEVICES=1,2 python train.py --accelerator gpu \ --devices 1 \ --max_epochs 500 \ --precision 16 \ --protein_num_layers 3 \ --ligand_num_layers 3 \ --complex_num_layers 3 \ --protein_node_h_dim 200 32 \ --protein_edge_h_dim 64 2 \ --ligand_node_h_dim 200 32 \ --ligand_edge_h_dim 64 2 \ --complex_node_h_dim 200 32 \ --complex_edge_h_dim 64 2 \ --dataset_name PDBBind \ --input_type multistage-physical \ --model_name gvp-multistage \ --data_dir /home/ec2-user/SageMaker/efs/data/PIGNet/data/pdbbind_v2019/scoring \ --residual \ --num_workers 8 \ --lr 1e-4 \ --bs 128 \ --early_stopping_patience 10 \ --default_root_dir /home/ec2-user/SageMaker/efs/model_logs/zichen/PDBBind_MSGVP_physical