Add new inplace api Tensor.uniform_, test=develop

[zoooo0820]

PR types

New features

PR changes

APIs

Describe

Add a new api Tensor.uniform_, which is the inplace version of api paddle.uniform.

Example:
code-block:: python

import paddle

x = paddle.ones(shape=[3, 4])
x.uniform_()
print(x)

# result is random
Tensor(shape=[3, 4], dtype=float32, place=CUDAPlace(0), stop_gradient=True,
       [[ 0.97134161, -0.36784279, -0.13951409, -0.48410338],
        [-0.15477282,  0.96190143, -0.05395842, -0.62789059],
        [-0.90525085,  0.63603556,  0.06997657, -0.16352385]])

[paddle-bot-old]

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

[CLAassistant]

All committers have signed the CLA.

[jerrywgz]

The random result is in a range of [min, max)

[zoooo0820]

done

[jerrywgz]

单测增加下输入为空的情况？

[zoooo0820]

Done,增加了维度含0的输入测试

[jzhang533]

这个是说seed设置成0的时候，用paddle.seed设置的全局随机数种子来生成随机数吧。
0 means use a seed generated by the system. 没表述清楚。

[zoooo0820]

done， 更新了seed参数的说明

[jzhang533]

LGTM

[jeff41404]

Please refer to other inplace operator's implementation, under normal situation, it can be shared with files of non-inplace, so no need to add uniform_random_inplace_op.cc, uniform_random_inplace_op.cu and uniform_random_inplace_op_xpu.cc

[zoooo0820]

Because the parameters between paddle.uniform and paddle.Tensor.uniform_ are different, non-inplace version Uniform Kernel can not be directly reused.

[jeff41404]

paddle can generate inplace function in Python, please refer to python/paddle/fluid/layers/ops.py and python/paddle/fluid/layers/layer_function_generator.py

[Superjomn]

LGTM

[XiaoguangHu01]

LGTM

[jzhang533]

lgtm