[wip] float8 training: invert the meaning of scale #1351
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
Summary: Needs some work before review. Benefit 1: less kernels for delayed scaling, so easier to debug things. Benefit 2: aligns with other tensor subclasses in torchao better. Test Plan: Some FSDP2 numerics tests are still failing, TODO debug Reviewers: Subscribers: Tasks: Tags:
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/1351
Note: Links to docs will display an error until the docs builds have been completed. ❌ 2 New FailuresAs of commit feec964 with merge base 6ff3904 ( NEW FAILURES - The following jobs have failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
vkuzo
commented
Nov 26, 2024
| @@ -34,8 +34,10 @@ def addmm_float8_unwrapped( | |||
| as inputs. This is used to standardize the logic between subclassed and non subclassed | |||
| versions of the linear module. | |||
| """ | |||
| a_inverse_scale = a_scale.reciprocal() | |||
| b_inverse_scale = b_scale.reciprocal() | |||
| # a_inverse_scale = a_scale.reciprocal() | |||
There was a problem hiding this comment.
not having to do this is what reduces the # of kernels for delayed scaling
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Outdated suggestions cannot be applied.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:
Needs some work before review.
Benefit 1: less kernels for delayed scaling, so easier to debug things.
Benefit 2: aligns with other tensor subclasses in torchao better.
Test Plan:
Some FSDP2 numerics tests are still failing, TODO debug
Reviewers:
Subscribers:
Tasks:
Tags: