Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

【Hackathon No.16】update RFC for RRelu #137

Merged
merged 6 commits into from May 24, 2022
Merged

【Hackathon No.16】update RFC for RRelu #137

Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
74cf458
add rfc for rrelu
thunder95 Mar 28, 2022
98e515d
modify date error
thunder95 Mar 28, 2022
5dfe766
rrelu训练阶段使用组合API
thunder95 Apr 1, 2022
ce3f965
保持文档一致性
thunder95 Apr 7, 2022
7ded244
Merge branch 'master' of https://github.com/PaddlePaddle/community in…
thunder95 May 24, 2022
3e84931
update rrelu rfc doc
thunder95 May 24, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
15 changes: 10 additions & 5 deletions rfcs/APIs/20220329_api_design_for_RRelu.md
View file
Open in desktop
@@ -1,6 +1,6 @@
# paddle.nn.RReLU设计文档
# paddle.nn.RReLU, paddle.nn.functional.rrelu设计文档

|API名称 | paddle.nn.RReLU |
|API名称 | paddle.nn.RReLU, paddle.nn.functional.rrelu |
|---|---|
|提交作者 | thunder95 |
|提交时间 | 2022-03-29 |
Expand Down Expand Up @@ -152,7 +152,14 @@ Pytorch支持inplace操作, 能有效降低显存占用。
API设计为`paddle.nn.RRelu(lower=1. / 8., upper=1. / 3., name=None)`，

- **lower** (float) - 均匀分布的下边界。
- **upper**(float) - 均匀分布的上边界。
- **upper** (float) - 均匀分布的上边界。

`paddle.nn.functional.rrelu(x, lower=1. / 8., upper=1. / 3., training=True, name=None)`,

- **x** (tensor) - 输入的张量。
- **lower** (float) - 均匀分布的下边界。
- **upper** (float) - 均匀分布的上边界。
- **training** (bool) - 是否在训练阶段, 默认是True。

## 底层OP设计
RRelu与飞桨中已有的PRelu在操作上是类似的，参考PReluOp设计来实现RRelu的底层OP。
Expand All @@ -163,8 +170,6 @@ RRelu与飞桨中已有的PRelu在操作上是类似的，参考PReluOp设计来

## 代码实现文件路径

在文件paddle/phi/kernels/impl/pixel_unshuffle_kernel_impl.h和paddle/phi/kernels/impl/pixel_unshuffle_grad_kernel_impl.h中编写主要的正向和反向计算逻辑：

CPU中正向和反向计算逻辑：
paddle/phi/kernels/cpu/rrelu_grad_kernel.cc
paddle/phi/kernels/cpu/rrelu_kernel.cc
Expand Down