def from_params(cls,
vocab: Vocabulary,
params: Params,
constructor_to_call=None,
constructor_to_inspect=None) -> 'BertModel':
#initialize the class using JSON params
embedder_params = params.pop("text_field_embedder")
token_params = embedder_params.pop("tokens")
embedding = PretrainedTransformerEmbedder.from_params(
vocab=vocab, params=token_params)
text_field_embedder = BasicTextFieldEmbedder(
token_embedders={'tokens': embedding})
# text_field_embedder = TextFieldEmbedder.from_params(vocab, embedder_params)
seq2vec_encoder_params = params.pop("seq2vec_encoder")
seq2vec_encoder = Seq2VecEncoder.from_params(seq2vec_encoder_params)
initializer = InitializerApplicator(
) #.from_params(params.pop("initializer", []))
params.assert_empty(cls.__name__)
# print(cls)
return cls(vocab=vocab,
text_field_embedder=text_field_embedder,
seq2vec_encoder=seq2vec_encoder,
initializer=initializer)
def test_forward_runs_when_initialized_from_params(self):
# This code just passes things off to pytorch-transformers, so we only have a very simple
# test.
params = Params({'model_name': 'bert-base-uncased'})
embedder = PretrainedTransformerEmbedder.from_params(params)
tensor = torch.randint(0, 100, (1, 4))
output = embedder(tensor)
assert tuple(output.size()) == (1, 4, 768)
def test_forward_runs_when_initialized_from_params(self):
# This code just passes things off to `transformers`, so we only have a very simple
# test.
params = Params({"model_name": "bert-base-uncased"})
embedder = PretrainedTransformerEmbedder.from_params(params)
token_ids = torch.randint(0, 100, (1, 4))
mask = torch.randint(0, 2, (1, 4))
output = embedder(token_ids=token_ids, mask=mask)
assert tuple(output.size()) == (1, 4, 768)