Please add "stop_gradient" parameter into paddle.create_parameter()

[ILoveAmy]

需求描述 Feature Description

To make the code more concise and elegant, Please add "stop_gradient" parameter into paddle.create_parameter()

import paddle
# 初始化权重参数，设置为不需要梯度
W_query = paddle.create_parameter(shape=[2, 6], dtype='float32', default_initializer=paddle.nn.initializer.Uniform(), stop_gradient=False)

Currently, we have to implement the code as below, TWO lines!.

import paddle
# 初始化权重参数，设置为不需要梯度
W_query = paddle.create_parameter(shape=[d_in, d_out], dtype='float32', default_initializer=paddle.nn.initializer.Uniform())
W_query.stop_gradient = True

However, We can implement it using only ONE line PyTorch code

W_query = torch.nn.Parameter(torch.rand(2, 6), requires_grad=False)

[ZhangHandi]

Thanks for your feedback. Currently, you need to use combination of code instead of implementation directly as mentioned in this tutorial create_parameter and I will feedback this issue to the corresponding developer.

[ILoveAmy]

Thanks for your feedback. Currently, you need to use combination of code instead of implementation directly as mentioned in this tutorial create_parameter and I will feedback this issue to the corresponding developer.

Thank you very much! Looking forward to Paddle's concise implementation of paddle.create_parameter()