Skip to content
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

Neo Quantization Fixes #2724

Merged
merged 2 commits into from
Feb 6, 2025
Merged

Conversation

a-ys
Copy link
Contributor

@a-ys a-ys commented Feb 5, 2025

Description

Various fixes for neo fp8 CI:

  • Update llm/prepare.py to properly pass in option for FP8 quantization.
  • Do tokenization from FP8 on our side, like we were doing before with AutoFP8. Delegating this to llmcompressor leads to unneeded tokenization that leads to timeout in CI.
  • Use lmi-dist-venv to run awq quantization, due to hf transformers version incompatibility.

Type of change

  • Bug fix (non-breaking change which fixes an issue)

Feature/Issue validation/testing

a-ys added 2 commits February 5, 2025 19:23
- for fp8, do dataset prep and tokenization on our side rather than
  through llmcompressor. prevents timeout.
- use lmi-dist venv for awq, due to incompatible hf transformers version
@a-ys a-ys requested review from zachgk and a team as code owners February 5, 2025 23:31
@siddvenk siddvenk merged commit 772f17f into deepjavalibrary:master Feb 6, 2025
9 checks passed
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.

2 participants