Skip to content

AITemplate 0.2

Latest
Latest
Compare
Choose a tag to compare
@ipiszy ipiszy released this 31 Jan 05:49
· 664 commits to main since this release
0.2
e160474
  • FX2AIT: A connector between Pytorch and AITemplate. Check more details in https://github.com/facebookincubator/AITemplate/tree/main/fx2ait/README.md. Another connector between PT2 and AITemplate is WIP, stay tuned.
  • Data type support: Added support for fp32. Bfloat16 is WIP.
  • Profiling: Reduced profiling time for both GeMM and Conv kernels.
  • Tooling: Added simple per-op profiling. Improved error messages.
  • Examples: The StableDiffusion example has been updated.
Assets 2
Loading