Remove FP16_Optimizer patch for DeepSpeed#2213
Merged
Merged
Conversation
Deepspeed already includes the same FusedAdam FP16_Optimizer originally from NVIDIA/apex here: https://github.com/deepspeedai/DeepSpeed/blob/master/deepspeed/runtime/fp16/fused_optimizer.py Currently this line gives a warning saying: ``` /opt/conda/envs/py_3.10/lib/python3.10/site-packages/onnxruntime/training/optim/_modifier_registry.py:56: UserWarning: Skip modifying optimizer because of optimizer name not found in the registry: accelerate.utils.deepspeed.DeepSpeedOptimizerWrapper ``` i.e. the FP16 Optimizer from onnxruntime (https://github.com/microsoft/onnxruntime/blob/main/orttraining/orttraining/python/training/optim/fp16_optimizer.py) is not actually wrapping the DeepSpeed fused Adam optimizer anyway, so this line is redundant.
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
IlyasMoutawwakil
approved these changes
Apr 2, 2025
Member
IlyasMoutawwakil
left a comment
IlyasMoutawwakil
left a comment
There was a problem hiding this comment.
LGTM, tests are passing !
is onnxruntime-training still useful for you ? we are discussing deprecating it since usage is so low.
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Deepspeed (since 2022) already includes the FusedAdam FP16_Optimizer originally from NVIDIA/apex here (and it's better maintained wrt deepspeed grad_norm, MOE training, etc): https://github.com/deepspeedai/DeepSpeed/blob/master/deepspeed/runtime/fp16/fused_optimizer.py
Currently this line gives a warning saying:
which is actually caused by
get_full_qualified_type_name(optimizer.optimizer) = 'deepspeed.runtime.fp16.fused_optimizer.FP16_Optimizer'not being in the onnxruntime registry: https://github.com/microsoft/onnxruntime/blob/main/orttraining/orttraining/python/training/optim/_modifier_registry.pyi.e. the FP16 Optimizer from onnxruntime (https://github.com/microsoft/onnxruntime/blob/main/orttraining/orttraining/python/training/optim/fp16_optimizer.py) is not actually doing anything, just falling back to the DeepSpeed fused Adam optimizer anyway, so this line and the warning it causes are redundant.
What does this PR do?
Fixes # (issue)
Before submitting
Who can review?
@JingyaHuang