kernl

Kernl lets you run PyTorch transformer models several times faster on GPU with a single line of code, and is designed to be easily hackable.

APACHE-2.0 License

Stars
1.5K

Bot releases are hidden (Show)

kernl - v0.2.2 Latest Release

Published by els-lab-ci over 1 year ago

0.2.2 (2023-02-17)

Bug Fixes

kernl - v0.2.1

Published by els-lab-ci over 1 year ago

0.2.1 (2023-02-08)

Bug Fixes

Documentation

  • add dynamo code to notebook as a comment (#274) (55cb686)
kernl - v0.2.0

Published by jonathlela over 1 year ago

New features

  • Whisper support
  • New debugger
  • Improve warmup speed
  • Improve CUDA graph memory allocation

What's Changed

New Contributors

Full Changelog: https://github.com/ELS-RD/kernl/compare/v0.1.0...v0.2.0

kernl - first release

Published by pommedeterresautee almost 2 years ago

  • covers Bert (and derivatives), and partially T5
  • 3 kernels: Flash Attention, LayerNorm, LinearLayer
  • debug mode
  • tutorials
Badges
Extracted from project README
Tests
Related Projects