num_train = len(dataset_train)
indices = list(range(num_train))
split = int(np.floor(args.valid_size * num_train))
train_idx, valid_idx = indices[split:], indices[:split]
valid_sampler = SubsetRandomSampler(valid_idx)
valid_loader = torch.utils.data.DataLoader(dataset_valid,
batch_size=args.batch_size,
num_workers=args.num_worker,
sampler=valid_sampler,
pin_memory=True)
controller = Controller().to(device)
child = Child().to(device)
controller.load_state_dict(torch.load('./save_model/controller.pt'))
child.load_state_dict(torch.load('./save_model/child.pt'))
best_reward = 0
for i in range(10):
controller.init_hidden(batch_size=1)
outputs = controller.sample_child()
normal_arc, reduction_arc, _, _ = outputs
for batch_idx, (inputs, targets) in enumerate(valid_loader):
outputs = controller.sample_child()
normal_arc, reduction_arc, entropy_seq, log_prob_seq = outputs
correct = 0
total = 0
# 1. get reward using a single mini-batch of validation data
inputs = inputs.to(device)
