Skip to content

Conversation

@legitnull
Copy link
Collaborator

@legitnull legitnull commented Jan 5, 2026

PR Category

Train

PR Types

Bug Fixes

PR Description

  • Add a flag use_amp to enable mixed precision training
  • Remove the accelerate dep, now the PI0/Pi0.5 training works with DDP
  • Remove the example pi0_old along with the model definition, train/infer/server script
  • Reorder the code structure
    • move transformation.diffusion to inference/core/diffusion
    • rename diffusion_entrypoint.py to inference_diffusion.py and move to inference/core

@legitnull legitnull changed the title Fix PI0 training with AMP Remove the accelerate dependency Jan 6, 2026
aoyulong
aoyulong previously approved these changes Jan 6, 2026
Copy link
Contributor

@aoyulong aoyulong left a comment

Choose a reason for hiding this comment

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

LGTM

@aoyulong aoyulong changed the title Remove the accelerate dependency Refactor the code and rm accelerate dependency Jan 7, 2026
@aoyulong aoyulong merged commit 3794714 into flagos-ai:main Jan 7, 2026
5 of 14 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.

4 participants