[Auto Parallel] add dropout spmd rules #70216
Merged
Conversation
This file contains hidden or 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
|
你的PR提交成功，感谢你对开源项目的贡献! |
jeff41404
reviewed
Dec 13, 2024
|
|
||
| // args : (Tensor x, Tensor seed_tensor, Scalar p, bool is_test, str mode, int | ||
| // seed, bool fix_seed) output : Tensor(out), Tensor(mask) | ||
| SpmdInfo DropoutFwdInferSpmd(const DistMetaTensor& x, |
There was a problem hiding this comment.
In general, the naming of forward spmd rule is DropoutInferSpmd
There was a problem hiding this comment.
由于现在不需要写reverse，所以命名为fwd和bwd更合理，已经有部分spmd是如此命名的了
jeff41404
reviewed
Dec 13, 2024
|
|
||
| // args : (Tensor mask, Tensor out_grad, Scalar p, bool is_test, str mode) | ||
| // output : Tensor(x_grad) | ||
| SpmdInfo DropoutBwdInferSpmd(const DistMetaTensor& mask, |
There was a problem hiding this comment.
In general, the naming of backward spmd rule is DropoutGradInferSpmd
jeff41404
reviewed
Dec 13, 2024
| bool is_test, | ||
| const std::string& mode) { | ||
| return ElementwiseBinaryInferSpmd(mask, out_grad); | ||
| } |
There was a problem hiding this comment.
shall we add DropoutInferSpmdReverse to support reverse spmd rule in forward when use static mode
There was a problem hiding this comment.
现在不需要添加reverse的切分推导了
zyfncg
approved these changes
Dec 16, 2024
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Outdated suggestions cannot be applied.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 Category
Auto Parallel
PR Types
Performance
Description
dropout切分推导规则添加
Pcard-73145
dropout切分推导会导致dp情况下精度无法诸位对齐，gpt中使用了dropout，因此gpt3个开启dp的单测会无法通过。暂时先禁止这3个测试，待此PR合入之后，再修改paddleNLP中gpt的测试。