NVIDIA / NVIDIA/TransformerEngine

[Feature Request] Integration of DiT components into TransformerEngine.

Open
#900 1 comment 2 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

enhancement
Dominant language
Python
Stars
3.5k
Forks
831
Avg merge
3d 11h
Merged PRs (30d)
65

Description

References

Two academic papers are cited to support the request:

Scalable Diffusion Models with Transformers: https://arxiv.org/abs/2212.09748
PixArt-α: Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis: https://arxiv.org/abs/2310.00426

Current Support

TransformerEngine does not yet support DiT.

Differences Noted

Specific differences such as LN elementwise_affine=False and Transformer layer with Time step aware scale / shift are highlighted.

Screenshot 2024-06-10 at 10 24 40

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

No files, tests, or entry points are named in the issue. Start by reading the cited DiT and PixArt-α papers and comparing their LayerNorm and time-step scale/shift requirements with TransformerEngine's current support; done means DiT components are integrated with those differences covered.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Quiet
Clarity
Needs clarification
Newbie friendliness
30/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.