def test_get_param_groups_for_optimizer(self):
word_embedding = WordEmbedding(
num_embeddings=5,
embedding_dim=4,
embeddings_weight=None,
init_range=(-1, 1),
unk_token_idx=4,
pad_token_idx=3,
mlp_layer_dims=[],
)
char_embedding = CharacterEmbedding(num_embeddings=5,
embed_dim=4,
out_channels=2,
kernel_sizes=[1, 2])
embedding_list = EmbeddingList([word_embedding, char_embedding],
concat=True)
param_groups = embedding_list.get_param_groups_for_optimizer()
self.assertEqual(len(param_groups), 1)
param_names = set(param_groups[0].keys())
expected_param_names = {
name
for name, _ in embedding_list.named_parameters()
}
self.assertSetEqual(param_names, expected_param_names)
class Config(RNNGParserBase.Config):
class ModelInput(BaseModel.Config.ModelInput):
tokens: TokenTensorizer.Config = TokenTensorizer.Config(
column="tokenized_text")
actions: AnnotationNumberizer.Config = AnnotationNumberizer.Config(
)
inputs: ModelInput = ModelInput()
embedding: WordEmbedding.Config = WordEmbedding.Config()
def test_generator(self):
model = Seq2SeqModel.from_config(
Seq2SeqModel.Config(
source_embedding=WordEmbedding.Config(embed_dim=512),
target_embedding=WordEmbedding.Config(embed_dim=512),
),
self._get_tensorizers(),
)
sample, _ = get_example_and_check()
repacked_inputs = {
"src_tokens": sample[0].t(),
"src_lengths": sample[1]
}
scripted_generator = ScriptedSequenceGenerator(
[model.model], model.trg_eos_index,
ScriptedSequenceGenerator.Config())
scripted_preds = scripted_generator.generate_hypo(repacked_inputs)
self.assertIsNotNone(scripted_preds)
class Config(BaseModel.Config):
class ModelInput(Model.Config.ModelInput):
tokens: TokenTensorizer.Config = TokenTensorizer.Config(
add_bos_token=True, add_eos_token=True)
inputs: ModelInput = ModelInput()
embedding: WordEmbedding.Config = WordEmbedding.Config()
representation: BiLSTM.Config = BiLSTM.Config(bidirectional=False)
decoder: Optional[MLPDecoder.Config] = MLPDecoder.Config()
output_layer: LMOutputLayer.Config = LMOutputLayer.Config()
tied_weights: bool = False
stateful: bool = False
def test_empty_mlp_layer_dims(self):
num_embeddings = 5
embedding_dim = 4
embedding_module = WordEmbedding(
num_embeddings=num_embeddings,
embedding_dim=embedding_dim,
embeddings_weight=None,
init_range=[-1, 1],
unk_token_idx=4,
mlp_layer_dims=[],
)
self.assertEqual(embedding_module.embedding_dim, embedding_dim)
def _create_dummy_model(self):
return create_model(
DocModel_Deprecated.Config(
representation=BiLSTMDocAttention.Config(
save_path=self.representation_path),
decoder=MLPDecoder.Config(save_path=self.decoder_path),
),
FeatureConfig(
word_feat=WordEmbedding.Config(
embed_dim=300, save_path=self.word_embedding_path),
save_path=self.embedding_path,
),
self._create_dummy_meta_data(),
)
class Config(BaseModel.Config):
class ModelInput(Model.Config.ModelInput):
tokens: Optional[TokenTensorizer.Config] = TokenTensorizer.Config(
add_bos_token=True, add_eos_token=True)
inputs: ModelInput = ModelInput()
embedding: WordEmbedding.Config = WordEmbedding.Config()
representation: Union[BiLSTM.Config,
CNN.Config] = BiLSTM.Config(bidirectional=False)
decoder: Optional[MLPDecoder.Config] = MLPDecoder.Config()
output_layer: LMOutputLayer.Config = LMOutputLayer.Config()
tied_weights: bool = False
stateful: bool = False
caffe2_format: ExporterType = ExporterType.PREDICTOR
def test_basic(self):
# Setup embedding
num_embeddings = 5
output_dim = 6
embedding_module = WordEmbedding(
num_embeddings=num_embeddings,
embedding_dim=4,
embeddings_weight=None,
init_range=[-1, 1],
unk_token_idx=4,
mlp_layer_dims=[3, output_dim],
)
self.assertEqual(embedding_module.embedding_dim, output_dim)
# Check output shape
input_batch_size, input_len = 4, 6
token_ids = torch.randint(
low=0, high=num_embeddings, size=[input_batch_size, input_len]
)
output_embedding = embedding_module(token_ids)
expected_output_dims = [input_batch_size, input_len, output_dim]
self.assertEqual(list(output_embedding.size()), expected_output_dims)
