class Config(NewDocModel.Config):
class RegressionModelInput(Model.Config.ModelInput):
tokens: TokenTensorizer.Config = TokenTensorizer.Config()
labels: NumericLabelTensorizer.Config = NumericLabelTensorizer.Config()
inputs: RegressionModelInput = RegressionModelInput()
output_layer: RegressionOutputLayer.Config = RegressionOutputLayer.Config()
def from_config(cls, config: Config, tensorizers: Dict[str, Tensorizer]):
embedding = cls.create_embedding(config, tensorizers)
representation = create_module(config.representation,
embed_dim=embedding.embedding_dim)
decoder = create_module(config.decoder,
in_dim=representation.representation_dim,
out_dim=1)
output_layer = RegressionOutputLayer.from_config(config.output_layer)
return cls(embedding, representation, decoder, output_layer)
class Config(NewBertModel.Config):
class InputConfig(ConfigBase):
tokens: BERTTensorizer.Config = BERTTensorizer.Config(
columns=["text1", "text2"], max_seq_len=128
)
labels: NumericLabelTensorizer.Config = NumericLabelTensorizer.Config()
inputs: InputConfig = InputConfig()
output_layer: RegressionOutputLayer.Config = RegressionOutputLayer.Config()
def from_config(cls, config: Config, tensorizers: Dict[str, Tensorizer]):
vocab = tensorizers["tokens"].vocab
encoder = create_module(
config.encoder,
padding_idx=vocab.get_pad_index(),
vocab_size=vocab.__len__(),
)
decoder = create_module(
config.decoder, in_dim=encoder.representation_dim, out_dim=1
)
output_layer = RegressionOutputLayer.from_config(config.output_layer)
return cls(encoder, decoder, output_layer)
