Fix autobatch size issue #1073
Merged
Fix autobatch size issue #1073
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
…items, plus updated the logic in generation size to respect what the user asks
clefourrier
commented
Nov 20, 2025
| if config.model_parallel is False and self.config.dtype not in ["4bit", "8bit"]: | ||
| logger.info(f"Using Data Parallelism, putting model on device {self._device}") | ||
| self.model = self.model.to(self._device) | ||
| if config.compile: |
Member
Author
There was a problem hiding this comment.
duplicate code, already exists in _create_auto_model
clefourrier
commented
Nov 20, 2025
| ) | ||
| # model.to(self.device) | ||
| model.eval() | ||
| torch.set_grad_enabled(False) |
Member
Author
There was a problem hiding this comment.
set at the module level
clefourrier
commented
Nov 20, 2025
| continuation = continuation.lstrip() | ||
| return continuation | ||
|
|
||
| def _model_call(self, inputs: torch.Tensor) -> torch.Tensor: |
Collaborator
|
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. |
Member
Author
|
@pcuenca this should fix the issue you had with autobatch size, can you take a look? I'm not sure it's 100% perfect, I'm still getting some memory not deallocated in the model, but I suspect it should already be helpful for your usecase |
NathanHB
approved these changes
Nov 20, 2025
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Outdated suggestions cannot be applied.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.
The PR does 2 main things:
delon the created objects to force the memory release of attached resourcesRest of modifs are nits (duplicated code/legacy functions that I removed), can put in another PR but they were thematically linked