fix gpt2 train loss Nan problem by add a line __syncthreads in BlockR… #33658
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
|
Thanks for your contribution! |
|
GPT train loss will NaN since the pr #33420 . |
jeff41404
reviewed
Jun 18, 2021
| d_scale_partial = BlockReduceSum<U>(d_scale_partial); | ||
| d_bias_partial = BlockReduceSum<U>(d_bias_partial); | ||
| __shared__ U shared_scale[32]; | ||
| __shared__ U shared_bias[32]; |
There was a problem hiding this comment.
用 kMaxBlockDim/lwarpSize 来代替32是否会更节省share memory
There was a problem hiding this comment.
已线下讨论，share memory整体空间足够大，可以保持32。
jeff41404
previously approved these changes
Jun 18, 2021
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Outdated suggestions cannot be applied.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
Bug fixes
PR changes
OPs
Describe
背景：
gpt2 训练过程中出现loss不稳定、不收敛、最终变成nan的情况。
经排查：
1）在P40上训练正常，V100上训练出现异常。
2）添加一行log打印训练正常，无log打印训练异常。
3）使用原线性相加方式训练正常，使用BlockReduceSum训练异常。
最终通过去掉static共享内存shared，同时添加一行__syncthreads后解决训练异常问题。
同时对另外两个BlockReduceSum加入__syncthreads以提高可靠性。