def test_tensor_matmul():
t1 = ts.Tensor([[1., 2., 3.], [4., 5., 6.]])
t2 = ts.Tensor([[7., 8.], [9., 10.], [11., 12.]])
t = t1 @ t2
expected_array = np.array([[58., 64.], [139., 154.]])
assert t.shape == expected_array.shape
np.testing.assert_equal(t.numpy, expected_array)
def test_tensor_add():
t1 = ts.Tensor([[1., 1.], [2., 2.]])
t2 = ts.Tensor([[2., 2.], [1., 1.]])
t = t1 + t2
expected_array = np.array([[3., 3.], [3., 3.]])
assert t.shape == expected_array.shape
np.testing.assert_equal(t.numpy, expected_array)
def test_matrix():
matrix = ts.Tensor([[1, 2, 3], [4, 5, 6]])
assert matrix.shape == (2, 3)
assert matrix[0, 0] == 1
assert matrix[0, 1] == 2
assert matrix[1, 0] == 4
assert matrix[1, 2] == 6
def test_log():
t1 = ts.Tensor([[2., 2.], [4., 4.]])
t = ts.log(t1)
expected_array = np.log(t1.numpy)
np.testing.assert_almost_equal(t.numpy, expected_array)
def test_tensor3():
tensor3 = ts.Tensor([[[1, -1], [2, -2], [3, -3]],
[[4, -4], [5, -5], [6, -6]]])
assert tensor3.shape == (2, 3, 2)
# assert tensor3[0] == [[1, -1], [2, -2], [3, -3]]
# assert tensor3[0, 2] == [3, -3]
assert tensor3[0, 2, 1] == -3
def test_exp():
t1 = ts.Tensor([[2., 2.], [4., 4.]])
t = ts.exp(t1)
expected_array = np.exp(t1.numpy)
np.testing.assert_almost_equal(t.numpy, expected_array, decimal=6)
def test_pow():
t1 = ts.Tensor([[2., 2.], [4., 4.]])
t = ts.pow(t1, 4)
expected_array = np.power(t1.numpy, 4)
np.testing.assert_almost_equal(t.numpy, expected_array)
def test_tensor_transpose():
tensor = ts.Tensor(np.array([[1., 2., 3.], [1., 2., 3.]]))
tensor_t = tensor.T
expected_array = np.array([[1., 1.], [2., 2.], [3., 3.]])
np.testing.assert_equal(tensor_t.numpy, expected_array)
def test_tensor_mul():
t1 = ts.Tensor([[1., 2.], [3., 4.]])
t = 2. * t1
expected_array = np.array([[2., 4.], [6., 8.]])
assert t.shape == expected_array.shape
np.testing.assert_equal(t.numpy, expected_array)
def test_tensor4():
tensor3 = ts.Tensor([[[[1, -1], [2, -2], [3, -3]],
[[4, -4], [5, -5], [6, -6]]],
[[[7, -7], [8, -8], [9, -9]],
[[10, -10], [11, -11], [12, -12]]]])
assert tensor3.shape == (2, 2, 3, 2)
# assert tensor3[1, 1, 1] == [11, -11]
assert tensor3[1, 1, 1, 1] == -11
def test_tensor_reshape():
tensor = ts.Tensor([
[[1., 2.], [3., 4.]],
[[1., 2.], [3., 4.]],
[[1., 2.], [3., 4.]],
])
t1 = tensor.reshape([3, 4])
expected_t1 = ts.Tensor([
[1., 2., 3., 4.],
[1., 2., 3., 4.],
[1., 2., 3., 4.],
])
assert t1.shape == expected_t1.shape
np.testing.assert_equal(t1.numpy, expected_t1.numpy)
t2 = tensor.reshape([3, 2, 2])
assert t2.shape == tensor.shape
np.testing.assert_equal(t2.numpy, tensor.numpy)
def test_scalar():
scalar = ts.Tensor(7)
assert scalar.shape == (1, )
assert scalar[0] == 7
def test_tensor_get_item():
tensor = ts.Tensor([[1, 2], [3, 4]])
assert tensor[0, 0] == 1
assert tensor[0, 1] == 2
assert tensor[1, 0] == 3
assert tensor[1, 1] == 4
def test_tensor_shape():
expected_shape = (2, 2)
tensor = ts.Tensor(shape=expected_shape)
assert isinstance(tensor.shape, tuple)
assert tensor.shape == expected_shape
def test_not_tensor_create_not_supported():
with pytest.raises(ValueError) as e:
ts.Tensor(shape=(1, 2, 3, 4, 5))
expected = "Tensor with dims higher than 4 are not supported yet! Note that len(shape)=5"
assert str(e.value) == expected
def test_vector():
vector = ts.Tensor([1, 2, 3])
assert vector.shape == (3, )
assert vector[1] == 2
def test_tensor_get_item_scalar():
tensor = ts.Tensor(7)
assert tensor[0] == 7
