sharding supports reduce_avg communication #62147
Merged
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
|
你的PR提交成功，感谢你对开源项目的贡献! |
ForFishes
reviewed
Mar 4, 2024
| @@ -55,6 +56,17 @@ def all_reduce(tensor, op=ReduceOp.SUM, group=None, sync_op=True): | |||
| >>> print(data) | |||
| >>> # [[5, 7, 9], [5, 7, 9]] (2 GPUs) | |||
| """ | |||
| # AVG is only supported when nccl >= 2.10 | |||
| if op == ReduceOp.AVG and paddle.base.core.nccl_version() < 21000: | |||
| tensor.scale_(1.0 / group.nranks) | |||
| self.use_reduce_avg = ( | ||
| strategy.hybrid_configs['sharding_configs'].use_reduce_avg | ||
| and paddle.base.core.nccl_version() >= 21000 | ||
| ) |
There was a problem hiding this comment.
开启了use_reduce_avg，实际nccl版本不符合，是不是需要有报错或者warning
ForFishes
reviewed
Mar 5, 2024
| # AVG is only supported when nccl >= 2.10 | ||
| if op == ReduceOp.AVG and paddle.base.core.nccl_version() < 21000: | ||
| group = ( | ||
| paddle.distributed.collective.new_group( |
There was a problem hiding this comment.
直接那global group吧，不需要新建一个group吧。
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Outdated suggestions cannot be applied.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 optimizationPR changes
OthersDescription
sharding supports reduce_avg communicationPcard-70426
收敛性测试，llama2-7b
中文文档：PaddlePaddle/docs#6514