_, step_loss, step_logits, step_attns = self.step(encoder_masks, img_data, zero_paddings, decoder_inputs, target_weights, bucket_id, self.forward_only)
curr_step_time = (time.time() - start_time)
step_time += curr_step_time / self.steps_per_checkpoint
logging.info("step_time: %f, loss: %f, step perplexity: %f"%(curr_step_time, step_loss, math.exp(step_loss) if step_loss < 300 else float("inf")))
loss += step_loss / self.steps_per_checkpoint
current_step += 1
step_outputs = [b for b in np.array([np.argmax(logit, axis=1).tolist() for logit in step_logits]).transpose()]
if self.visualize:
After Change
total = (self.s_gen.get_size() // self.batch_size)
with tqdm(desc="Train: ", total=total) as pbar:
for epoch in range(self.num_epoch):
logging.info("Generating first batch)")
for i,batch in enumerate(self.s_gen.gen(self.batch_size)):
//logging.info("Generated batch number: {}".format(i))
// Get a batch and make a step.
