def create_and_check_model(self, config, input_ids, token_type_ids,
sequence_labels, token_labels, choice_labels):
model = FNetModel(config=config)
model.to(torch_device)
model.eval()
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_inference_model(self):
model = FNetModel.from_pretrained("google/fnet-base")
model.to(torch_device)
input_ids = torch.tensor([[0, 1, 2, 3, 4, 5]], device=torch_device)
output = model(input_ids)[0]
expected_shape = torch.Size((1, 6, model.config.hidden_size))
self.assertEqual(output.shape, expected_shape)
expected_slice = torch.tensor(
[[[4.1541, -0.1051, -0.1667], [-0.9144, 0.2939, -0.0086],
[-0.8472, -0.7281, 0.0256]]],
device=torch_device)
self.assertTrue(
torch.allclose(output[:, :3, :3], expected_slice, atol=1e-4))
def test_model_from_pretrained(self):
for model_name in FNET_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
model = FNetModel.from_pretrained(model_name)
self.assertIsNotNone(model)