CV-CUDA icon indicating copy to clipboard operation
CV-CUDA copied to clipboard

[QUESTION]Python 3.13t Support

Open pangkun248 opened this issue 2 months ago • 2 comments

Is there any support plan available?

pangkun248 avatar Oct 27 '25 02:10 pangkun248

nvImageCodec also supports up to Python 3.13t: https://github.com/NVIDIA/nvImageCodec/tree/main

pangkun248 avatar Oct 27 '25 02:10 pangkun248

Hi @pangkun248 , thank you for your interest in CV-CUDA!

We are investigating support for the free-threaded Python versions, but we do not have a timeline on when this will be available.

Do you have more context on your application requiring support for the free-threaded Python versions? If you have additional context to provide it can help us prioritize our work.

justincdavis avatar Nov 14 '25 18:11 justincdavis

I want to use CV-CUDA to accelerate preprocessing and post-processing operations in my video stream inference service. I want to decouple preprocessing, inference, and post-processing into different processes/threads for operation. However, if multiple processes are used, there will be significant delay in data transmission between them. Starting from Python 3.13t, the GIL restriction has been lifted, so I want to try this method.

pangkun248 avatar Nov 17 '25 17:11 pangkun248