def test_zeros():
"""Tests zeros"""
pf.set_backend("pytorch")
# Scalar
zeros = ops.zeros([1])
assert isinstance(zeros, torch.Tensor)
assert zeros.ndim == 1
assert zeros.shape[0] == 1
assert zeros.numpy() == 0.0
# 1D
zeros = ops.zeros([5])
assert isinstance(zeros, torch.Tensor)
assert zeros.ndim == 1
assert zeros.shape[0] == 5
assert all(zeros.numpy() == 0.0)
# 2D
zeros = ops.zeros([5, 4])
assert isinstance(zeros, torch.Tensor)
assert zeros.ndim == 2
assert zeros.shape[0] == 5
assert zeros.shape[1] == 4
assert np.all(zeros.numpy() == 0.0)
# 3D
zeros = ops.zeros([5, 4, 3])
assert isinstance(zeros, torch.Tensor)
assert zeros.ndim == 3
assert zeros.shape[0] == 5
assert zeros.shape[1] == 4
assert zeros.shape[2] == 3
assert np.all(zeros.numpy() == 0.0)
def test_mean():
"""Tests mean"""
pf.set_backend('pytorch')
# Should mean along the last dimension by default
ones = torch.ones([5, 4, 3])
val = ops.mean(ones)
assert isinstance(val, torch.Tensor)
assert val.ndim == 2
assert val.shape[0] == 5
assert val.shape[1] == 4
assert np.all(val.numpy() == 1.0)
# But can change that w/ the axis kwarg
ones = torch.ones([5, 4, 3])
val = ops.mean(ones, axis=1)
assert isinstance(val, torch.Tensor)
assert val.ndim == 2
assert val.shape[0] == 5
assert val.shape[1] == 3
assert np.all(val.numpy() == 1.0)
# Actually test values
val = ops.mean(torch.Tensor([0.9, 1.9, 2.1, 3.1]))
assert is_close(val.numpy(), 2.0)
def test_add_col_of():
"""Tests add_col_of"""
pf.set_backend('pytorch')
a = torch.randn([2, 3, 5])
val = ops.add_col_of(a, 1)
assert val.ndim == 3
assert val.shape[0] == 2
assert val.shape[1] == 3
assert val.shape[2] == 6
def test_kl_divergence():
"""Tests kl_divergence"""
pf.set_backend("pytorch")
# Divergence between a distribution and itself should be 0
dist = torch.distributions.normal.Normal(0, 1)
assert ops.kl_divergence(dist, dist).numpy() == 0.0
# Divergence between two different distributions should be >0
d1 = torch.distributions.normal.Normal(0, 1)
d2 = torch.distributions.normal.Normal(1, 1)
assert ops.kl_divergence(d1, d2).numpy() > 0.0
# Divergence between more different distributions should be larger
d1 = torch.distributions.normal.Normal(0, 1)
d2 = torch.distributions.normal.Normal(1, 1)
d3 = torch.distributions.normal.Normal(2, 1)
assert (ops.kl_divergence(d1, d2).numpy() < ops.kl_divergence(d1,
d3).numpy())
# Should auto-convert probflow distibutions
dist = pf.Normal(0, 1)
assert ops.kl_divergence(dist, dist).numpy() == 0.0
