test_loader = DataLoader(test_data, batch_size=args.batch_size, shuffle=True)
print('Finish loading the data....')
if not args.aligned:
print("### Note: You are running in unaligned mode.")
####################################################################
#
# Hyperparameters
#
####################################################################
hyp_params = args
hyp_params.orig_d_l, hyp_params.orig_d_a, hyp_params.orig_d_v = train_data.get_dim(
)
hyp_params.l_len, hyp_params.a_len, hyp_params.v_len = train_data.get_seq_len()
hyp_params.layers = args.nlevels
hyp_params.use_cuda = use_cuda
hyp_params.dataset = dataset
hyp_params.when = args.when
hyp_params.batch_chunk = args.batch_chunk
hyp_params.n_train, hyp_params.n_valid, hyp_params.n_test = len(
train_data), len(valid_data), len(test_data)
hyp_params.model = str.upper(args.model.strip())
hyp_params.output_dim = output_dim_dict.get(dataset, 1)
hyp_params.criterion = criterion_dict.get(dataset, 'L1Loss')
if __name__ == '__main__':
test_loss = train.initiate(hyp_params, train_loader, valid_loader,
test_loader)
def main():
params = load_config(args)
params, train_loader, valid_loader, test_loader = load_data_pipeline(params)
save_config(params)
test_loss = train.initiate(params, train_loader, valid_loader, test_loader)