def test_warning_torch_and_autograd(self):
"""Test that a warning is raised if the sequence of tensors contains
both torch and autograd tensors."""
x = torch.tensor([1.0, 2.0, 3.0])
y = np.array([0.5, 0.1])
with pytest.warns(UserWarning, match="Consider replacing Autograd with vanilla NumPy"):
fn._get_multi_tensorbox([x, y])
def test_exception_tensorflow_and_torch(self):
"""Test that an exception is raised if the sequence of tensors contains
tensors from incompatible dispatch libraries"""
x = tf.Variable([1.0, 2.0, 3.0])
y = onp.array([0.5, 0.1])
z = torch.tensor([0.6])
with pytest.raises(ValueError, match="Tensors contain mixed types"):
fn._get_multi_tensorbox([x, y, z])
def test_return_numpy_box(self):
"""Test that NumPy is correctly identified as the dispatching library."""
x = onp.array([1.0, 2.0, 3.0])
y = [0.5, 0.1]
res = fn._get_multi_tensorbox([y, x])
assert res.interface == "numpy"
def test_return_torch_box(self):
"""Test that Torch is correctly identified as the dispatching library."""
x = torch.tensor([1.0, 2.0, 3.0])
y = onp.array([0.5, 0.1])
res = fn._get_multi_tensorbox([y, x])
assert res.interface == "torch"
def test_return_tensorflow_box(self):
"""Test that TensorFlow is correctly identified as the dispatching library."""
x = tf.Variable([1.0, 2.0, 3.0])
y = onp.array([0.5, 0.1])
res = fn._get_multi_tensorbox([y, x])
assert res.interface == "tf"
