Add CUDA 12.9 build

[matthewdouglas]

• Add CUDA Toolkit 12.9 builds for Windows and Linux platforms.
• Update test suite to test against nightly PyTorch 2.8.0+cu129

Resolves #1688, #1642

[github-actions]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[FurkanGozukara]

i am still getting this error what could be reason?

2025-10-01 23:23:28 WARNING Could not find the bitsandbytes CUDA binary at PosixPath('/workspace/kohya_ss/venv/lib/python3.10/site-packages/bitsandbytes/libbitsandbytes_cuda129.so') cextension.py:89
WARNING The installed version of bitsandbytes was compiled without GPU support. 8-bit optimizers, 8-bit multiplication, and GPU quantization are unavailable.

[matthewdouglas]

@FurkanGozukara You're likely using an older version of bitsandbytes. This PR was first included in v0.47.0.

pip install -U bitsandbytes

If you still need help please open a separate issue.

[FurkanGozukara]

@matthewdouglas thanks my bad i think the repo had auto installed 0.44 not 0.48