def setUp(self):
layer_dim = 512
xnmt.events.clear()
ParamManager.init_param_col()
self.model = DefaultTranslator(
src_reader=PlainTextReader(),
trg_reader=PlainTextReader(),
src_embedder=SimpleWordEmbedder(emb_dim=layer_dim, vocab_size=100),
encoder=BiLSTMSeqTransducer(input_dim=layer_dim,
hidden_dim=layer_dim),
attender=MlpAttender(input_dim=layer_dim,
state_dim=layer_dim,
hidden_dim=layer_dim),
trg_embedder=SimpleWordEmbedder(emb_dim=layer_dim, vocab_size=100),
decoder=AutoRegressiveDecoder(
input_dim=layer_dim,
trg_embed_dim=layer_dim,
rnn=UniLSTMSeqTransducer(input_dim=layer_dim,
hidden_dim=layer_dim,
decoder_input_dim=layer_dim,
yaml_path="model.decoder.rnn"),
transform=NonLinear(input_dim=layer_dim * 2,
output_dim=layer_dim),
scorer=Softmax(input_dim=layer_dim, vocab_size=100),
bridge=CopyBridge(dec_dim=layer_dim, dec_layers=1)),
)
self.model.set_train(False)
self.src_data = list(
self.model.src_reader.read_sents("examples/data/head.ja"))
self.trg_data = list(
self.model.trg_reader.read_sents("examples/data/head.en"))
self.search = GreedySearch()
def test_single(self):
dy.renew_cg()
train_loss = self.model.calc_loss(
src=self.src_data[0],
trg=self.trg_data[0],
loss_calculator=AutoRegressiveMLELoss()).value()
dy.renew_cg()
outputs = self.model.generate(
xnmt.batcher.mark_as_batch([self.src_data[0]]), [0],
GreedySearch(),
forced_trg_ids=xnmt.batcher.mark_as_batch([self.trg_data[0]]))
output_score = outputs[0].score
self.assertAlmostEqual(-output_score, train_loss, places=5)
def test_single(self):
dy.renew_cg()
train_loss = self.model.calc_loss(src=self.src_data[0],
trg=self.trg_data[0],
loss_calculator=MLELoss()).value()
dy.renew_cg()
self.model.initialize_generator()
outputs = self.model.generate_output(self.src_data[0],
0,
GreedySearch(),
forced_trg_ids=self.trg_data[0])
output_score = outputs[0].score
self.assertAlmostEqual(-output_score, train_loss, places=5)
def initialize_generator(self, **kwargs):
if kwargs.get("len_norm_type", None) is None:
len_norm = xnmt.length_normalization.NoNormalization()
else:
len_norm = xnmt.serializer.YamlSerializer().initialize_if_needed(kwargs["len_norm_type"])
search_args = {}
if kwargs.get("max_len", None) is not None: search_args["max_len"] = kwargs["max_len"]
if kwargs.get("beam", None) is None:
self.search_strategy = GreedySearch(**search_args)
else:
search_args["beam_size"] = kwargs.get("beam", 1)
search_args["len_norm"] = len_norm
self.search_strategy = BeamSearch(**search_args)
self.report_path = kwargs.get("report_path", None)
self.report_type = kwargs.get("report_type", None)
def test_greedy_vs_beam(self):
dy.renew_cg()
outputs = self.model.generate(
xnmt.batcher.mark_as_batch([self.src_data[0]]), [0],
BeamSearch(beam_size=1),
forced_trg_ids=xnmt.batcher.mark_as_batch([self.trg_data[0]]))
output_score1 = outputs[0].score
dy.renew_cg()
outputs = self.model.generate(
xnmt.batcher.mark_as_batch([self.src_data[0]]), [0],
GreedySearch(),
forced_trg_ids=xnmt.batcher.mark_as_batch([self.trg_data[0]]))
output_score2 = outputs[0].score
self.assertAlmostEqual(output_score1, output_score2)
def initialize_generator(self, **kwargs):
if kwargs.get("len_norm_type", None) is None:
len_norm = xnmt.length_normalization.NoNormalization()
else:
len_norm = initialize_object(kwargs["len_norm_type"])
search_args = {}
if kwargs.get("max_len", None) is not None:
search_args["max_len"] = kwargs["max_len"]
self.max_len = kwargs.get("max_len", 50)
if kwargs.get("beam", None) is None:
self.search_strategy = GreedySearch(**search_args)
else:
search_args["beam_size"] = kwargs.get("beam", 1)
search_args["len_norm"] = len_norm
# self.search_strategy = TransformerBeamSearch(**search_args)
self.report_path = kwargs.get("report_path", None)
self.report_type = kwargs.get("report_type", None)
def test_greedy_vs_beam(self):
dy.renew_cg()
self.model.initialize_generator()
outputs = self.model.generate_output(self.src_data[0],
0,
BeamSearch(beam_size=1),
forced_trg_ids=self.trg_data[0])
output_score1 = outputs[0].score
dy.renew_cg()
self.model.initialize_generator()
outputs = self.model.generate_output(self.src_data[0],
0,
GreedySearch(),
forced_trg_ids=self.trg_data[0])
output_score2 = outputs[0].score
self.assertAlmostEqual(output_score1, output_score2)