-
Notifications
You must be signed in to change notification settings - Fork 216
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
mx formats: create MXLinearConfig #1688
base: main
Are you sure you want to change the base?
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/1688
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 1064e83 with merge base d3306b2 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
Summary: Creating a config to make it easier to enable more configuration options. This is important as we enable more functionality from pytorch/pytorch#146414 in torchao. Note that future PRs will flesh out the config, to keep PR size small. Test Plan: ``` pytest test/prototype/mx_formats/ -s -x ``` Reviewers: Subscribers: Tasks: Tags: ghstack-source-id: d6b63187f87dee1467ba2a6377863023df7d9f6f ghstack-comment-id: 2648821215 Pull Request resolved: #1688
x_mx_dq = to_dtype( | ||
x_mx._data, | ||
x_mx._scale_e8m0, | ||
x_mx._elem_dtype, | ||
x_mx._block_size, | ||
hp_dtype, # noqa: E501 | ||
use_fp4_custom_triton_dequant_kernel, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why is this needed? I did a quick pass through the RFC and didn't see any reference to this functionality
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
we'll probably delete it in a future PR. The current PR is a refactor with no logic changes.
Summary:
Creating a config to make it easier to enable more configuration
options. This is important as we enable more functionality from
pytorch/pytorch#146414 in torchao.
Note that future PRs will flesh out the config, to keep PR size small.
Test Plan:
Reviewers:
Subscribers:
Tasks:
Tags: