from eICU_preprocessing.split_train_test import create_folder
from models.run_tpc import TPC
from models.initialise_arguments import initialise_tpc_arguments
if __name__ == '__main__':
c = initialise_tpc_arguments()
c['mode'] = 'test'
c['exp_name'] = 'TempWeightShare'
c['model_type'] = 'temp_only'
c['share_weights'] = True
c['temp_kernels'] = [32] * c['n_layers']
log_folder_path = create_folder('models/experiments/final', c.exp_name)
temp_weight_share = TPC(
config=c,
n_epochs=c.n_epochs,
name=c.exp_name,
base_dir=log_folder_path,
explogger_kwargs={'folder_format': '%Y-%m-%d_%H%M%S{run_number}'})
temp_weight_share.run()
from models.run_tpc import TPC
import numpy as np
from models.initialise_arguments import initialise_tpc_arguments
from captum.attr import IntegratedGradients
from torch import cat, ones
time_point = 23 # this means 24 hours because of python's indexing
batch_size = 2
if __name__ == '__main__':
base_dir = 'models/experiments/final/TPC'
exp_dir = 'models/experiments/final/TPC/2020-06-27_2048171'
tpc = TPC(base_dir=base_dir,
resume=exp_dir,
resume_save_types=('model', 'simple', 'th_vars', 'results'))
c = initialise_tpc_arguments()
c['mode'] = 'test'
c['exp_name'] = 'TPC'
c['model_type'] = 'tpc'
tpc.config = c
tpc.setup()
tpc._setup_internal()
tpc.prepare()
tpc.model.eval()
ig = IntegratedGradients(tpc.model)