NVIDIA / NVIDIA/Megatron-LM

[BUG]--cross_entropy_fusion_impl te results in grad_norm 0

Open
#1,750 4 comments 0 reactions 0 assignees View on GitHub
bug module: transformer engine
Dominant language
Python
Stars
17.9k
Forks
4.5k
Avg merge
4d 6h
Merged PRs (30d)
271

Description

After passing through this function:
https://github.com/NVIDIA/TransformerEngine/blob/6ba98d439190901fec85c2ae3c2cea235d9f2196/transformer_engine/pytorch/triton/cross_entropy.py#L359-L365

_input becomes all zeros.

Image

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.