[Build] 1.19.0 still depends on CUDA 11.x

[tibortakacs]

Describe the issue

Although the description of the 1.19.0 release writes that "Default GPU packages use CUDA 12.x and Cudnn 9.x", the 1.19.0 build still depends on CUDA 11.x.

Urgency

No response

Target platform

Windows

Build script

n/a

Error / output

n/a

Visual Studio Version

No response

GCC / Compiler Version

No response

[Alarmod]

Work normal with CUDA 12.4.0 (with GPU) and don't work with CUDA 12.4.1 (dll files changed in CUDA 12.4.1)

[tianleiwu]

It's true that both onnxruntime-win-x64-gpu-1.19.0.zip and onnxruntime-win-x64-cuda-1.19.0.zip depends on cuda 11.x and cudnn8.x. A walkaround is to copy the dll files from onnxruntime-gpu 1.19.0 python package, which supports cuda 12.x and cudnn 9.x.

The coming 1.19.2 release shall support cuda 12.x and cudnn 9.x in onnxruntime-win-x64-gpu-1.19.2.zip.

[tianleiwu]

@tibortakacs, Please try onnxruntime-win-x64-gpu-1.19.2.zip in https://github.com/microsoft/onnxruntime/releases/tag/v1.19.2

[Alarmod]

@tibortakacs, Please try onnxruntime-win-x64-gpu-1.19.2.zip in https://github.com/microsoft/onnxruntime/releases/tag/v1.19.2
Tested with CUDA 12.4.0 + CUDNN 9.2.0 and 9.2.1 - working OK

[tibortakacs]

Thank you, all, for hard work and fixing this. Fantastic.

[juntaosun]

@tibortakacs, Please try onnxruntime-win-x64-gpu-1.19.2.zip in https://github.com/microsoft/onnxruntime/releases/tag/v1.19.2

When releasing the pip onnxruntime-gpu package, can 11.x also be added, for example:

19.2+cu118 , 19.2 ,  19.0+cu118,  19.0,

When installing in this way, you only need to:

pip install onnxruntime-gpu==19.2+cu118

or cuda 12.x and cudnn 9.x

pip install onnxruntime-gpu==19.2

@tianleiwu

[tianleiwu]

@juntaosun, you can install python package for cuda 11.8 (and cudnn 8.9) like

pip install onnxruntime-gpu==1.19.2 --extra-index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/onnxruntime-cuda-11/pypi/simple/