NVIDIA / NVIDIA/TransformerEngine
Heuristic Fallback in cuDNN: Why THD Format Reverts to SM80 on Hopper GPUs
Open
@cyanguwa is already working on this.
Since Feb 3, 2026.
question
- Dominant language
- Python
- Stars
- 3.5k
- Forks
- 831
- Avg merge
- 3d 11h
- Merged PRs (30d)
- 65
Description
"Why does the forward pass call the cuDNN SM90 interface when the learnable_sink option is enabled with qkv_format="bshd", but reverts to the cuDNN SM80 interface when qkv_format="thd" is selected?"
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Assessment
This issue has not been assessed yet.