for i in range(number_of_batches):
op_start_time = time.time()
batch_x, batch_y = data.train.next_batch(batch_size)
model.train_on_batch(
batch_x.reshape(-1, img_height, img_width, channels), batch_y)
# Log (log :)) loss, current position and times
op_time, h, m, s = get_times(op_start_time, start_time)
#if (i + 1) % 50 == 0:
loss, acc = model.evaluate(
batch_x.reshape(-1, img_height, img_width, channels), batch_y,
batch_size)
lr = get_lr(i + 1)
logger.color_print(
logger.Bold,
'epoch {} | batch {}/{} | loss: {:.2f} | acc: {} | lr: {:.4E} | {:.2f}s | {:02d}:{:02d}:{:02d}'
.format(epoch + 1, i + 1, number_of_batches, loss, acc, lr,
op_time, h, m, s))
#else:
# print('epoch {} | batch {}/{} | {:.2f}s | {:02d}:{:02d}:{:02d}'
# .format(epoch + 1, i + 1, number_of_batches, op_time, h, m, s))
# Approximate train log score with ~1/4 dataset size for efficiency
evaluate(data.train, data.train.get_number_of_batches(batch_size))
evaluate(data.valid, data.valid.get_number_of_batches(batch_size))
score_data['lr'] += [get_lr(data.train.get_number_of_batches(batch_size))]
#score_data['f1_score'] += [f1_score(...)]
#scores_path = scores_template.format(epoch, batch_size, opt_name, lr_starting, lr_decay)
#print('Saving scores (train/valid loss, lr and f1-score) to {}'.format(scores_path))
for i in range(number_of_batches):
op_start_time = time.time()
batch_x, batch_y = data.train.next_batch(batch_size)
model.train_on_batch(
batch_x.reshape(-1, img_height, img_width, channels), batch_y)
# Log (log :)) loss, current position and times
op_time, overall_h, overall_m, overall_s = get_times(
op_start_time, start_time)
if (i + 1) % 50 == 0:
loss = model.evaluate(
batch_x.reshape(-1, img_height, img_width, channels), batch_y,
batch_size)
logger.color_print(
logger.Info,
'epoch {0} | batch {1} / {2} | loss: {3:.2f} | {4:.2f}s | {5:02d}:{6:02d}:{7:02d}'
.format(epoch + 1, i + 1, number_of_batches, loss, op_time,
overall_h, overall_m, overall_s))
else:
print(
'epoch {0} | batch {1} / {2} | {3:.2f}s | {4:02d}:{5:02d}:{6:02d}'
.format(epoch + 1, i + 1, number_of_batches, op_time,
overall_h, overall_m, overall_s))
# Approximate train log score with ~1/4 dataset size for efficiency
#log_score(data.train, iter_limit=data.train.get_number_of_batches(batch_size) // 4)
#log_score(data.valid)
current_lr = lr_starting * (
lr_decay**data.train.get_number_of_batches(batch_size))**(epoch + 1)
logger.color_print(logger.Info,
'Current learning rate: {}'.format(current_lr))
