fix mx triton kernel after PyTorch triton pin change #431
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/431
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ✅ No FailuresAs of commit 24dd661 with merge base 37c348e ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
Closed
|
That was fast, for a test in CI you'd need to revert the changes from here as well #428 |
vkuzo
force-pushed
the
20240624_mx_bugfix
branch
from
bc234a8 to
3330c07
Compare
Summary: Triton pin updated recently: pytorch/pytorch#126098 In the new triton version, functions can only access global variables of type `tl.constexpr`. Due to the current structure of the code and the fact that these constants are also used by non-triton programs, I think the best thing to do is to just stop using globals in the MX triton kernel. The PR lifts all of these constants to kernel function arguments. Test Plan: ``` pytest test/prototype/mx_formats/test_custom_cast.py ``` Reviewers: Subscribers: Tasks: Tags:
vkuzo
force-pushed
the
20240624_mx_bugfix
branch
from
3330c07 to
24dd661
Compare
msaroufim
approved these changes
Jun 24, 2024
yanbing-j
pushed a commit
to yanbing-j/ao
that referenced
this pull request
Dec 9, 2024
Unless user specifies device, default it to CUDA if it's available on the platform. As for all models, other than TinyLlama inference on GPU will be faster than on CPU
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:
Triton pin updated recently:
pytorch/pytorch#126098
In the new triton version, functions can only access global variables of type
tl.constexpr. Due to the current structure of the code and the fact that these constants are also used by non-triton programs, I think the best thing to do is to just stop using globals in the MX triton kernel. The PR lifts all of these constants to kernel function arguments.Test Plan:
Reviewers:
Subscribers:
Tasks:
Tags: