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
1.61k stars 256 forks source link

[Common] Added Alignment Requirements for CuBLAS heuristics #845

Closed phu0ngng closed 1 month ago

phu0ngng commented 1 month ago

Type of change

Checklist:

phu0ngng commented 1 month ago

/te-ci pytorch

ptrendx commented 1 month ago

Could we add some test for it?

deepakn94 commented 1 month ago

LGTM, thanks. I verified that my failing case now works.

phu0ngng commented 1 month ago

Hi, thanks for the feedback. Glad to hear that it works for you now @deepakn94. @ptrendx I added the unit test to capture this bug. Let me know if there is an issue. Thanks

phu0ngng commented 1 month ago

/te-ci pytorch