requirements/cuda.txt

-r common.txt
onnxruntime>=1.23.2
fa3-fwd==0.0.2
# nvidia-cublas-cu12 is pinned to 12.9.1.4 via force-reinstall in Dockerfile.ci
# because torch pins its own cublas version, making it unsatisfiable at resolve time.
[Hardware] [Feat] Setup platform dependent package installation (#1046 ) 2026-02-05 21:47:41 +08:00			`-r common.txt`
			`onnxruntime>=1.23.2`
Update pyproject.toml to remove redundant transformers package installation and adjust Dockerfile.ci to install the current package in editable mode. Also, update fa3-fwd package version in requirements/cuda.txt to 0.0.2 for consistency. 2026-02-11 21:18:53 +08:00			`fa3-fwd==0.0.2`
Update requirements/cuda.txt to clarify the pinning of nvidia-cublas-cu12 version 12.9.1.4 due to dependency resolution issues with torch. This comment aids in understanding the force-reinstall strategy in Dockerfile.ci. 2026-02-11 22:44:11 +08:00			`# nvidia-cublas-cu12 is pinned to 12.9.1.4 via force-reinstall in Dockerfile.ci`
			`# because torch pins its own cublas version, making it unsatisfiable at resolve time.`