def convert_orig_tf1_checkpoint_to_pytorch(tf_checkpoint_path,
convbert_config_file,
pytorch_dump_path):
conf = ConvBertConfig.from_json_file(convbert_config_file)
model = ConvBertModel(conf)
model = load_tf_weights_in_convbert(model, conf, tf_checkpoint_path)
model.save_pretrained(pytorch_dump_path)
def test_inference_no_head(self):
model = ConvBertModel.from_pretrained("YituTech/conv-bert-base")
input_ids = torch.tensor([[1, 2, 3, 4, 5, 6]])
output = model(input_ids)[0]
expected_shape = torch.Size((1, 6, 768))
self.assertEqual(output.shape, expected_shape)
expected_slice = torch.tensor([[[-0.0864, -0.4898, -0.3677],
[0.1434, -0.2952, -0.7640],
[-0.0112, -0.4432, -0.5432]]])
self.assertTrue(
torch.allclose(output[:, :3, :3], expected_slice, atol=1e-4))
def test_inference_masked_lm(self):
model = ConvBertModel.from_pretrained("YituTech/conv-bert-base")
input_ids = torch.tensor([[0, 1, 2, 3, 4, 5]])
output = model(input_ids)[0]
print(output[:, :3, :3])
expected_shape = torch.Size((1, 6, 768))
self.assertEqual(output.shape, expected_shape)
# TODO Replace values below with what was printed above.
expected_slice = torch.tensor(
[[[-0.0348, -0.4686, -0.3064], [0.2264, -0.2699, -0.7423], [0.1032, -0.4501, -0.5828]]]
)
self.assertTrue(torch.allclose(output[:, :3, :3], expected_slice, atol=1e-4))
def create_and_check_model(
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
):
model = ConvBertModel(config=config)
model.to(torch_device)
model.eval()
result = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids)
result = model(input_ids, token_type_ids=token_type_ids)
result = model(input_ids)
self.parent.assertEqual(result.last_hidden_state.shape, (self.batch_size, self.seq_length, self.hidden_size))
def test_model_from_pretrained(self):
for model_name in CONVBERT_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
model = ConvBertModel.from_pretrained(model_name)
self.assertIsNotNone(model)