fix typo (0d635085) · Commits · kotzaneck / IRL Machine Translation Project

joeynmt/model.py

+0 −1

Original line number	Diff line number	Diff line
		@@ -88,7 +88,6 @@ class Model(nn.Module):
		return encoder_output, encoder_hidden, None, None

		elif return_type == "decode":

		outputs, hidden, att_probs, att_vectors = self._decode(
		trg_input=kwargs["trg_input"],
		encoder_output=kwargs["encoder_output"],

joeynmt/prediction.py

+11 −3

Original line number	Diff line number	Diff line
		@@ -204,7 +204,10 @@ def test(cfg_file,
		cfg = load_config(cfg_file)

		if len(logger.handlers) == 0:
		make_logger(f'{cfg["training"]["model_dir"]}/test.log')
		log_file = None
		if os.path.exists(cfg["training"]["model_dir"]):
		log_file = f'{cfg["training"]["model_dir"]}/test.log'
		make_logger(log_file)

		if "test" not in cfg["data"].keys():
		raise ValueError("Test data must be specified in config.")
		@@ -282,7 +285,9 @@ def test(cfg_file,
		if data_set is None:
		continue

		logger.info(f"Decoding on {data_set_name} set...")
		dataset_filepath = cfg["data"][data_set_name] + "." + cfg["data"]["trg"]
		logger.info(f"Decoding on {data_set_name} set ({dataset_filepath})...")

		#pylint: disable=unused-variable
		score, loss, ppl, sources, sources_raw, references, hypotheses, \
		hypotheses_raw, attention_scores = validate_on_data(
		@@ -380,7 +385,10 @@ def translate(cfg_file, ckpt: str, output_path: str = None) -> None:
		cfg = load_config(cfg_file)

		#logger = make_logger()
		make_logger(f'{cfg["training"]["model_dir"]}/translation.log')
		log_file = None
		if os.path.exists(cfg["training"]["model_dir"]):
		log_file = f'{cfg["training"]["model_dir"]}/translation.log'
		make_logger(log_file)

		# when checkpoint is not specified, take oldest from model dir
		if ckpt is None:

joeynmt/training.py

+8 −8

Original line number	Diff line number	Diff line
		@@ -179,10 +179,6 @@ class TrainManager:
		self.model, self.optimizer = amp.initialize(self.model, self.optimizer,
		opt_level=fp16_opt_level)

		# multi-gpu training (should be after apex fp16 initialization)
		if self.n_gpu > 1:
		self.model = _DataParallel(self.model)

		# initialize accumalted batch loss (needed for batch_multiplier)
		#self.norm_batch_loss_accumulated = 0
		# initialize training statistics
		@@ -209,6 +205,10 @@ class TrainManager:
		reset_scheduler=reset_scheduler,
		reset_optimizer=reset_optimizer)

		# multi-gpu training (should be after apex fp16 initialization)
		if self.n_gpu > 1:
		self.model = _DataParallel(self.model)

		def _save_checkpoint(self) -> None:
		"""
		Save the model's current parameters and the training state to a
		@@ -306,7 +306,7 @@ class TrainManager:

		# fp16
		if self.fp16 and model_checkpoint.get("amp_state", None) is not None:
		amp.load_state_dict(checkpoint['amp_state'])
		amp.load_state_dict(model_checkpoint['amp_state'])

		# pylint: disable=unnecessary-comprehension
		# pylint: disable=too-many-branches
		@@ -335,14 +335,14 @@ class TrainManager:
		#
		# # gradient accumulation:
		# # loss.backward() inside _train_step()
		# epoch_loss += self._train_step(inputs)
		# batch_loss += self._train_step(inputs)
		#
		# if (i + 1) % self.batch_multiplier == 0:
		# self.optimizer.step() # update!
		# self.model.zero_grad() # reset gradients
		# self.steps += 1 # increment counter
		#
		# epoch_loss += batch_loss # add batch loss
		# epoch_loss += batch_loss # accumulate batch loss
		# batch_loss = 0 # reset batch loss
		#
		# # leftovers are just ignored.
		@@ -411,7 +411,7 @@ class TrainManager:
		elapsed = time.time() - start - total_valid_duration
		elapsed_tokens = self.total_tokens - start_tokens
		logger.info(
		"Epoch %3d Step: %8d Batch Loss: %12.6f "
		"Epoch %3d, Step: %8d, Batch Loss: %12.6f, "
		"Tokens per Sec: %8.0f, Lr: %.6f",
		epoch_no + 1, self.steps, batch_loss,
		elapsed_tokens / elapsed,

test/unit/test_batch.py

+7 −7

Original line number	Diff line number	Diff line
		@@ -69,9 +69,9 @@ class TestData(TensorTestCase):
		b = Batch(torch_batch=b, pad_index=self.pad_index)
		if total_samples == 0:
		self.assertTensorEqual(b.src, expected_src0)
		self.assertTensorEqual(b.src_lengths, expected_src0_len)
		self.assertTensorEqual(b.src_length, expected_src0_len)
		self.assertTensorEqual(b.trg, expected_trg0)
		self.assertTensorEqual(b.trg_lengths, expected_trg0_len)
		self.assertTensorEqual(b.trg_length, expected_trg0_len)
		total_samples += b.nseqs
		self.assertLessEqual(b.nseqs, batch_size)
		self.assertEqual(total_samples, len(self.train_data))
		@@ -117,18 +117,18 @@ class TestData(TensorTestCase):

		# test the sorting by src length
		self.assertEqual(type(b), Batch)
		before_sort = b.src_lengths
		b.sort_by_src_lengths()
		after_sort = b.src_lengths
		before_sort = b.src_length
		b.sort_by_src_length()
		after_sort = b.src_length
		self.assertTensorEqual(torch.sort(before_sort, descending=True)[0],
		after_sort)
		self.assertEqual(type(b), Batch)

		if total_samples == 0:
		self.assertTensorEqual(b.src, expected_src0)
		self.assertTensorEqual(b.src_lengths, expected_src0_len)
		self.assertTensorEqual(b.src_length, expected_src0_len)
		self.assertTensorEqual(b.trg, expected_trg0)
		self.assertTensorEqual(b.trg_lengths, expected_trg0_len)
		self.assertTensorEqual(b.trg_length, expected_trg0_len)
		total_samples += b.nseqs
		self.assertLessEqual(b.nseqs, batch_size)
		self.assertEqual(total_samples, len(self.dev_data))

test/unit/test_metric.py

+1 −1

Original line number	Diff line number	Diff line
		@@ -34,7 +34,7 @@ class TestMetrics(TensorTestCase):
		ref = ["あれがテストです。"]
		score = bleu(hyp, ref, tokenize="ja-mecab")
		self.assertAlmostEqual(score, 39.764, places=3)
		except ModuleNotFoundError as e:
		except Exception as e:
		raise unittest.SkipTest(f"{e} Skip.")

		def test_token_acc_level_char(self):