dont let dynamo inline inside of NF4 constructors or __torch_dispatch__#544
Merged
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/544
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 7b10019 with merge base afde175 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
Hanxian97
pushed a commit
that referenced
this pull request
Jul 29, 2024
…__ (#544) * dont let dynamo inline inside of NF4 constructors or __torch_dispatch__ * allow_in_graph the ctr instead of disabling it
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
This fixes a few known "gotchas" for
NF4Tensorwhen getting it to work with compile. Enumerating:(1) the NF4Tensor() constructor needs to be wrapped in a function that is annotated with
torch._dynamo.allow_in_graph. We want to make this work more automatically soon, see tracking issue: pytorch/pytorch#114389(2)
NF4Tensor.__new__andNF4Tensor.__torch_dispatch__both need to be marked withtorch._dynamo.disable.There isn't really a great fix for (2) today. One thing to note though is that as long as your model doesn't have a million graph breaks, you are unlikely to run into problems with (2). In the torchtune repro, there were a lot of deeply nested graph breaks in the FSDP2 code that caused dynamo to try to e.g. install itself directly onto the
NF4Tensor.__new__frame, which causes problems.