【Hackathon No.32】为 Paddle 优化 expand_as 前向&反向 op 在 GPU 上的计算性能 #52700
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Co-authored-by: Timber-Ye <ye_hanqiao@163.com> Co-authored-by: BrianQian1999 <brianqianhitsz@gmail.com>
update expand_as_perf
|
你的PR提交成功，感谢你对开源项目的贡献! |
|
Hi, we need TPM approval to pass the static test, so to complete the CI. @jzhang533 @wanglun @lileding @Superjomn |
zhangbopd
reviewed
Apr 10, 2023
| self.assertTrue( | ||
| out_purefp16_fp32.dtype == fluid.core.VarDesc.VarType.FP32 | ||
| ) | ||
| pass |
There was a problem hiding this comment.
这里的assertion是因为原先的expand_as算子不支持fp16类型，我们添加了fp16支持后即可去掉此处的断言；
There was a problem hiding this comment.
这块如果注册fp16报错应该找到原因并进行修改，而不是直接把报错全部注释掉
|
@Timber-Ye @BrianQian1999 经过黑客松组委会讨论：
|
Done. |
zhangbopd
reviewed
Apr 13, 2023
| #include "paddle/phi/kernels/impl/expand_as_grad_kernel_impl.h" | ||
| #include "paddle/phi/kernels/funcs/reduce_function.h" | ||
|
|
||
| #define MAX_RANK_SUPPORTED 6 |
zhangbopd
reviewed
Apr 13, 2023
| PADDLE_ENFORCE_LE( | ||
| out_rank, | ||
| MAX_RANK_SUPPORTED, | ||
| errors::InvalidArgument("The rank of the input 'Out@GRAD' for " |
There was a problem hiding this comment.
请问哪里不通顺呢？这里直接借鉴的是文件expand_grad_kernel_impl.h L94-L101。
JamesLim-sy
approved these changes
Apr 17, 2023
jjyaoao
pushed a commit
to jjyaoao/Paddle
that referenced
this pull request
Apr 19, 2023
…addle#52700) * Implement optimized kernel for OP-expand_as. * Support fp16. Co-authored-by: Timber-Ye <ye_hanqiao@163.com> Co-authored-by: BrianQian1999 <brianqianhitsz@gmail.com> * remove fp16 support * remove MAX_RANK_SUPPORTED --------- Co-authored-by: BrianQian1999 <brianqianhitsz@gmail.com>
|
hi, @Timber-Ye
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
PR types
Performance optimization
PR changes
OPs
Describe
目前 Paddle 内
expand_as前向和反向算子的 GPU 实现采用 Eigen 组合的模式，缺少 GPU Kernel，性能相对不足，希望实现高性能的 GPU 计算 Kernel，为 Paddle 优化expand_asop 在 GPU 上的计算性能。【算子性能优化设计文档】
由于expand_as前向的过程与广播机制类似，后向的过程与求和约归类似，因此直接通过使用飞桨内部的
BroadcastKernel和ReduceKernel来对expand_as算子进行优化。完成优化后，Paddle(Optimized)与优化前的Paddle(Baseline)的性能对比:
针对以上9种不同的case, 优化后的性能有所提升，并且要扩展的Tensor元素数量越多，性能提升越明显，优化后的算子在case 8上的用时更是直接缩短至baseline的1/814。
原始PR：