Fix the MoE-params gradient-scaling #4957
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
RezaYazdaniAminabadi
requested review from
tjruwase and
mrwyattii
as code owners
|
@tjruwase, can we merge this please? thanks :) |
tjruwase
approved these changes
Jan 17, 2024
github-merge-queue
bot
removed this pull request from the merge queue due to failed status checks
|
@mrwyattii can you please approve the workflows here? Thanks |
|
kind ping on this @tjruwase @mrwyattii |
mauryaavinash95
pushed a commit
to mauryaavinash95/DeepSpeed
that referenced
this pull request
Feb 17, 2024
This PR fixes a bug that I introduced in a previous [PR](microsoft#4695). The MoE-Params' gradients got accidentally double-scaled due to passing `self.ipg_bucket_has_moe_params` to the all_reduce functions. Since, we have already done the scaling the MoE parameters [here](https://github.com/microsoft/DeepSpeed/blob/master/deepspeed/runtime/zero/stage_1_and_2.py#L1054), we can safely pass `divide=False`. The divide argument may not be needed anymore, however, I just let it be there as I think it may be needed for the sequence-parallelism accuracy stability adjustments. cc: @tjruwase
rraminen
pushed a commit
to ROCm/DeepSpeed
that referenced
this pull request
May 9, 2024
This PR fixes a bug that I introduced in a previous [PR](microsoft#4695). The MoE-Params' gradients got accidentally double-scaled due to passing `self.ipg_bucket_has_moe_params` to the all_reduce functions. Since, we have already done the scaling the MoE parameters [here](https://github.com/microsoft/DeepSpeed/blob/master/deepspeed/runtime/zero/stage_1_and_2.py#L1054), we can safely pass `divide=False`. The divide argument may not be needed anymore, however, I just let it be there as I think it may be needed for the sequence-parallelism accuracy stability adjustments. cc: @tjruwase
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.
This PR fixes a bug that I introduced in a previous PR. The MoE-Params' gradients got accidentally double-scaled due to passing
self.ipg_bucket_has_moe_paramsto the all_reduce functions. Since, we have already done the scaling the MoE parameters here, we can safely passdivide=False. The divide argument may not be needed anymore, however, I just let it be there as I think it may be needed for the sequence-parallelism accuracy stability adjustments.cc: @tjruwase