fill_constant op supports NaN and Inf

[zhiqiu]

PR types

Bug fixes

PR changes

OPs

Describe

fill_constant op supports NaN and Inf

For example,

import paddle
import numpy as np
print(paddle.fluid.layers.fill_constant([1], 'float32', np.nan))
print(paddle.fluid.layers.fill_constant([1], 'float32', np.inf))

• before

Tensor(shape=[1], dtype=float32, place=CPUPlace, stop_gradient=True,
       [0.])
Tensor(shape=[1], dtype=float32, place=CPUPlace, stop_gradient=True,
       [0.])

• after

Tensor(shape=[1], dtype=float32, place=CPUPlace, stop_gradient=True,
       [nan.])
Tensor(shape=[1], dtype=float32, place=CPUPlace, stop_gradient=True,
       [inf.])

[paddle-bot-old]

Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

[wawltor]

LGTM