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
1.65k stars 261 forks source link

When ub_overlap_rs_dgrad is set to True, the error "Caught signal 8 (Floating point exception: integer divide by zero)" is raised. #788

Open JJGSBGQ opened 3 months ago

JJGSBGQ commented 3 months ago

Setting ub_overlap_rs_dgrad to True in megatron-LM will raise "Caught signal 8 (Floating point exception: integer divide by zero) "error, which was eventually found to be caused by a problem with the tex.gemm calculation in the backward.

196DF0A7-6E1C-43f2-B3E9-86FE4AEA76D3

image

JJGSBGQ commented 3 months ago

@minitu

Ethan-yt commented 1 month ago

+1