NVIDIA / TransformerEngine

A library for accelerating Transformer models on NVIDIA GPUs, including using 8-bit floating point (FP8) precision on Hopper and Ada GPUs, to provide better performance with lower memory utilization in both training and inference.
https://docs.nvidia.com/deeplearning/transformer-engine/user-guide/index.html
Apache License 2.0
1.6k stars 255 forks source link

[PyTorch] Disable THD test on architectures lower than sm90 #971

Closed cyanguwa closed 6 days ago

cyanguwa commented 6 days ago

Description

This PR disables CP tests for THD format for architectures lower than sm90. In the future, we will add a utility function so users can check this before running any tests.

Type of change

Changes

Please list the changes introduced in this PR:

Checklist:

cyanguwa commented 6 days ago

/te-ci pytorch

timmoon10 commented 6 days ago

Closed in favor of https://github.com/NVIDIA/TransformerEngine/pull/973.