[Misc] Move DP for ViT code inside model executor dir #25459
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
Signed-off-by: DarkLight1337 <[email protected]>
DarkLight1337
requested review from
NickLucche and
sighingnow
as code owners
DarkLight1337
added
the
ready
mergify
bot
added
llama
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request is a good refactoring that moves data parallelism utilities for vision models from vllm/multimodal/utils.py to a more appropriate location at vllm/model_executor/models/vision.py. The changes are well-contained, with tests moved alongside the code and all call sites updated accordingly. I have one suggestion to improve efficiency by creating a tensor on the correct device directly. Overall, this is a solid improvement to the codebase structure.
Comment on lines
+326
to
+327
| image_embeds_local = vision_model( | ||
| pixel_values_local, torch.tensor(local_grid_thw_list)) |
There was a problem hiding this comment.
For efficiency, it's better to create the local_grid_thw_list tensor directly on the target device to avoid an unnecessary CPU-to-GPU copy. The pixel_values_local tensor is already on the correct device.
Suggested change
| image_embeds_local = vision_model( | |
| pixel_values_local, torch.tensor(local_grid_thw_list)) | |
| image_embeds_local = vision_model( | |
| pixel_values_local, | |
| torch.tensor(local_grid_thw_list, | |
| device=pixel_values_local.device)) |
Isotr0py
approved these changes
Sep 23, 2025
FeiDaLI
pushed a commit
to FeiDaLI/vllm
that referenced
this pull request
Sep 25, 2025
…5459) Signed-off-by: DarkLight1337 <[email protected]>
charlifu
pushed a commit
to ROCm/vllm
that referenced
this pull request
Sep 25, 2025
…5459) Signed-off-by: DarkLight1337 <[email protected]> Signed-off-by: charlifu <[email protected]>
yewentao256
pushed a commit
that referenced
this pull request
Oct 3, 2025
Signed-off-by: DarkLight1337 <[email protected]> Signed-off-by: yewentao256 <[email protected]>
gjc0824
pushed a commit
to gjc0824/vllm
that referenced
this pull request
Oct 10, 2025
…5459) Signed-off-by: DarkLight1337 <[email protected]> Signed-off-by: gaojc <[email protected]>
xuebwang-amd
pushed a commit
to xuebwang-amd/vllm
that referenced
this pull request
Oct 10, 2025
…5459) Signed-off-by: DarkLight1337 <[email protected]> Signed-off-by: xuebwang-amd <[email protected]>
choprahetarth
pushed a commit
to Tandemn-Labs/vllm
that referenced
this pull request
Oct 11, 2025
…5459) Signed-off-by: DarkLight1337 <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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.
Purpose
Code cleanup: move the code from
multimodal/utils.pytomodel_executor/models/vision.pyTest Plan
Test Result
Essential Elements of an Effective PR Description Checklist
supported_models.mdandexamplesfor a new model.