Open
Description
Project URL
https://pypi.org/project/cutensor-cu13/
Does this project already exist?
- Yes
New Limit
400 MiB
Update issue title
- I have updated the title.
Which indexes
PyPI
About the project
This is a follow-up of #5643. NVIDIA releases CUDA 13, and this project is to support CUDA 13 users to use cuTENSOR. Like cutensor-cu12 we're requesting the same size limit for cutensor-cu13. Separation of CUDA 12/13 wheels is necessary as CUDA only guarantees minor version compatibility, not major version. We will drop support for cutensor-cu11 moving forward.
Reasons for the request
Similar to cutensor-cu12, adding support for the NVIDIA Blackwell architecture that was just released takes us over the 200MB limit. Currently, we expect wheels to be around ~250MB, and 400MB would give us breathing room to support our python users while we also investigate ways to reduce binary size.
Code of Conduct
- I agree to follow the PSF Code of Conduct