fmha: Support for different sequence length ranges

Open
#1,201 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
5/5
Estimated time
Over a week
Newbie friendliness
25/100
Issue type
Feature
Clarity
Needs clarification
Activity status
Stale
Tech stack
python, pytorch
Domain
machine-learning

Research direction

The issue points to fmha_kernel.h and the current 512-token limit; start there to locate the sequence-length and head/hidden-size constraints. Done means extending FMHA to support the requested ranges or configurations, including the wav2vec 2.0 use case, with verification of the resulting support.

Written by the indexing model from the issue text.

Description

Thanks a lot for the implementation of fmha!

Is it possible to extend the range of supported sequence lengths (currently its only up to 512) (or maybe different Head / Hidden size configurations) they will be super useful, for example to be used with huge models such as wav2vec 2.0.

Is the limitation defined in the fmha_kernel.h file?

Dominant language
Python
Stars
9k
Forks
1.5k
Avg merge
2d 4h
Merged PRs (30d)
3

Contributor guide

No contributing guide indexed for this repository

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.

More from NVIDIA/apex

All issues in NVIDIA/apex

Similar issues

More Python issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.