def test_bcast_semantics():
inputs = [
paddle.to_tensor(
np.ones(shape=[1, 3, 1, 1], dtype='float32', name="x9")),
paddle.to_tensor(
np.ones(shape=[1, 8, 1, 1], dtype='float32', name="x10"))
]
paddle.broadcast_tensors(inputs)
def test_type():
inputs = [
paddle.to_tensor(
np.ones(shape=[1, 1, 1, 1], dtype='float32', name="x4")),
paddle.to_tensor(
np.ones(shape=[1, 4, 1, 1], dtype='float64', name="x5"))
]
paddle.broadcast_tensors(inputs)
def test_dtype():
inputs = [
paddle.to_tensor(
np.ones(shape=[1, 1, 1, 1], dtype='int8', name="x6")),
paddle.to_tensor(
np.ones(shape=[1, 4, 1, 1], dtype='int8', name="x7"))
]
paddle.broadcast_tensors(inputs)
def test_dtype():
inputs = [
paddle.fluid.layers.data(shape=[1, 1, 1, 1],
dtype='int8',
name="x6"),
paddle.fluid.layers.data(shape=[1, 4, 1, 1],
dtype='int8',
name="x7")
]
paddle.broadcast_tensors(inputs)
def test_bcast_semantics():
inputs = [
paddle.fluid.layers.data(shape=[1, 3, 1, 1],
dtype='float32',
name="x9"),
paddle.fluid.layers.data(shape=[1, 8, 1, 1],
dtype='float32',
name="x10")
]
paddle.broadcast_tensors(inputs)
def test_type():
inputs = [
paddle.fluid.layers.data(shape=[1, 1, 1, 1],
dtype='float32',
name="x4"),
paddle.fluid.layers.data(shape=[1, 4, 1, 1],
dtype='float64',
name="x5")
]
paddle.broadcast_tensors(inputs)
def test_static():
inputs = [
paddle.fluid.layers.data(shape=[4, 1, 4, 1],
dtype='float32',
name="x0"),
paddle.fluid.layers.data(shape=[1, 4, 1, 4],
dtype='float32',
name="x1")
]
paddle.broadcast_tensors(inputs)
def test_dynamic():
paddle.disable_static()
try:
inputs = [
paddle.to_tensor(
np.random.random([4, 1, 4, 1]).astype("float32")),
paddle.to_tensor(
np.random.random([1, 4, 1, 4]).astype("float32"))
]
paddle.broadcast_tensors(inputs)
finally:
paddle.enable_static()
def __init__(self, alpha, beta):
if isinstance(alpha, numbers.Real):
alpha = paddle.full(shape=[1], fill_value=alpha)
if isinstance(beta, numbers.Real):
beta = paddle.full(shape=[1], fill_value=beta)
self.alpha, self.beta = paddle.broadcast_tensors([alpha, beta])
self._dirichlet = Dirichlet(paddle.stack([self.alpha, self.beta], -1))
super(Beta, self).__init__(self._dirichlet._batch_shape)
def log_prob(self, value):
"""probability mass function evaluated at value
Args:
value (Tensor): value to be evaluated.
Returns:
Tensor: probability of value.
"""
if paddle.is_integer(value):
value = paddle.cast(value, self.probs.dtype)
logits, value = paddle.broadcast_tensors(
[paddle.log(self.probs), value])
logits[(value == 0) & (paddle.isinf(logits))] = 0
return (paddle.lgamma(value.sum(-1) + 1) -
paddle.lgamma(value + 1).sum(-1) + (value * logits).sum(-1))
