val_loss.append(loss.item())
val_acc.append(acc.item())
postfix_dict = {"val_loss": loss.item(), "val_acc": acc.item()}
val_trange.set_postfix(**postfix_dict)
avg_loss = np.mean(val_loss[-600:])
avg_acc = np.mean(val_acc[-600:])
postfix = ' (Best)' if avg_acc >= best_acc else ' (Best: {})'.format(
best_acc)
print('Avg Val Loss: {}, Avg Val Acc: {}{}'.format(avg_loss, avg_acc,
postfix))
if avg_acc >= best_acc:
best_acc = avg_acc
best_model = copy.deepcopy(model.state_dict())
best_generator = copy.deepcopy(generator.state_dict())
if epoch % 5 == 0 and best_model != None:
torch.save(best_model, model_path)
torch.save(best_generator, generator_path)
print("model saved! - {}".format(model_path))
# init sampler
train_sampler = BatchSampler('train', train_dataset.labels, N_way,
sample_per_class, episodes)
train_dataloader = DataLoader(train_dataset,
batch_size=batch_size,
sampler=train_sampler)
# val_sampler = BatchSampler('val', val_dataset.labels, val_N_way, val_sample_per_class, episodes)
# val_dataloader = DataLoader(
device=opts.device)
val_loss.append(loss.item())
val_acc.append(acc.item())
avg_loss = np.mean(val_loss[-opts.iterations:])
avg_acc = np.mean(val_acc[-opts.iterations:])
postfix = ' (Best)' if avg_acc >= best_acc else ' (Best: {:.5f})'.format(
best_acc)
print_log(
'Avg Val Loss: {:.5f}, Avg Val Acc: {:.5f}{}'.format(
avg_loss, avg_acc, postfix), opts.log_file)
if avg_acc >= best_acc:
best_acc = avg_acc
if opts.multi_gpu:
torch.save(net.module.state_dict(), opts.model_file)
else:
torch.save(net.state_dict(), opts.model_file)
print_log('[epoch {} / iter {}] best model saved to: {}'.format(
epoch, len(train_db), opts.model_file),
file=opts.log_file)
best_acc = avg_acc
best_state = net.state_dict()
# TRAINING ENDS
if best_state is not None:
net.load_state_dict(best_state) # fixme when multi gpu, net.module()
print_log('Testing with best model ...', opts.log_file)
avg_acc = list()
for epoch in range(10):
test_iter = iter(test_db)
for batch in test_iter:
x, y = batch