Bugfix: Do not rely on model.device if model could be partially loaded
#7529
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
RyanJDick
requested review from
lstein,
blessedcoolant,
brandonrising,
hipsterusername and
psychedelicious
as code owners
github-actions
bot
added
python
psychedelicious
approved these changes
Jan 7, 2025
hipsterusername
approved these changes
Jan 8, 2025
hipsterusername
force-pushed
the
ryan/do-no-rely-on-model-device
branch
from
5d36c1c to
163b04e
Compare
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Outdated suggestions cannot be applied.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.
Summary
Given the recent introduction of partial model loading, the meaning of
model.devicehas changed. It could return the"cpu"device even if the model is partially loaded and should be passed a"cuda"input. This PR makes a best effort to go through and clean up places where we accessmodel.device.Related Issues / Discussions
Original bug report: https://discord.com/channels/1020123559063990373/1149506274971631688/1326245764258861117
QA Instructions
Smoke test all of the modified code paths:
I was also able to reproduce the error reported in Discord by manually loading models to the same level as OP (in a debugger). I confirmed that this changes fixed that particular issue.
Checklist
What's Newcopy (if doing a release after this PR)