def test_to_np(): array = np.arange(0, 3).astype(np.float) assert core.to_np(array) is array tensor = core.T(array) result = core.to_np([tensor, tensor]) np.testing.assert_equal(result[0], array) np.testing.assert_equal(result[1], array) variable = core.V(array) np.testing.assert_equal(core.to_np(variable), array)
def test_to_np(): array = np.arange(0, 3).astype(np.float) assert core.to_np(array) is array tensor = core.T(array) result = core.to_np([tensor, tensor]) np.testing.assert_equal(result[0], array) np.testing.assert_equal(result[1], array) variable = core.V(array) np.testing.assert_equal(core.to_np(variable), array) with mock.patch("torch.cuda") as cuda_mock: tensor_long = core.T(array.astype(np.int)) cuda_mock.is_available(return_value=True) cuda_mock.HalfTensor = torch.LongTensor array = core.to_np(tensor_long) np.testing.assert_equal(array, [0., 1., 2.]) assert array.dtype in (np.float32, np.float64)
def test_to_np(): array = np.arange(0, 3).astype(np.float) assert core.to_np(array) is array tensor = core.T(array) result = core.to_np([tensor, tensor]) np.testing.assert_equal(result[0], array) np.testing.assert_equal(result[1], array) variable = core.V(array) np.testing.assert_equal(core.to_np(variable), array) with mock.patch("torch.cuda.is_available") as is_available_mock: with mock.patch("fastai.core.is_half_tensor") as is_half_tensor_mock: is_available_mock.return_value = True is_half_tensor_mock.return_value = True tensor = core.T(array.astype(np.int)) array = core.to_np(tensor) np.testing.assert_equal(array, [0., 1., 2.]) assert array.dtype in (np.float32, np.float64)
def test_V(map_over_mock): core.V("foo") assert map_over_mock.call_args[0][0] == 'foo' assert type(map_over_mock.call_args[0][1]) == type(lambda: 0)
def test_forward(self): x = core.V(np.array([[1., 2.]]), requires_grad=False) output = core.to_np(self.simple_net.forward(x)) np.testing.assert_almost_equal(output, [[-1.435481, -0.27181]], decimal=4)