Skip to content

Attach act_max_hook for FP8 model#1447

Merged
yiliu30 merged 3 commits intomainfrom
fix-fp8-model
Feb 12, 2026
Merged

Attach act_max_hook for FP8 model#1447
yiliu30 merged 3 commits intomainfrom
fix-fp8-model

Conversation

@yiliu30
Copy link
Contributor

@yiliu30 yiliu30 commented Feb 12, 2026

Description

Fix #1390

Type of Change

  • Bug fix
  • New feature
  • Documentation update
  • Performance improvement
  • Code refactoring
  • Other (please specify):

Related Issues

Fixes or relates to #

Checklist Before Submitting

  • My code has been tested locally.
  • Documentation has been updated as needed.
  • New or updated tests are included where applicable.

Signed-off-by: yiliu30 <yi4.liu@intel.com>
Signed-off-by: yiliu30 <yi4.liu@intel.com>
Copilot AI review requested due to automatic review settings February 12, 2026 07:35
Copy link
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

This PR attaches an activation-maximum hook during blockwise RTN quantization, intended to support FP8 workflows.

Changes:

  • Registers act_max_hook on each quantization block during _quantize_via_rtn_blockwise.

@yiliu30 yiliu30 requested review from n1ck-guo and xin3he February 12, 2026 08:00
Copy link
Contributor

@lkk12014402 lkk12014402 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@yiliu30 yiliu30 merged commit 81c8ee4 into main Feb 12, 2026
29 checks passed
@yiliu30 yiliu30 deleted the fix-fp8-model branch February 12, 2026 08:54
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[Bug]: All experts of MiniMaxAI/MiniMax‑M2.1 are uncalibrated

2 participants