Installation process of Windows git version of StoryToolkitAI 0.24.0 reinstalls non-Cuda version of Torch
Closed this issue · 3 comments
Describe the bug
Installation process of Windows git version of StoryToolkitAI 0.24.0 reinstalls non-Cuda version of Torch. This makes StoryToolkitAI 0.24 cannot use NVIDIA card even if one is available.
To Reproduce
Steps to reproduce the behavior:
- Follow the installation process of Windows git version of StoryToolkitAI 0.24.0 including the installation of Torch with CUDA 2.2.1
cu118 - Execute venv\Scripts\python StoryToolkitAI\storytoolkitai 1st time
- You can see the installation of storytookitai removes Torch with CUDA 2.21 cu118 and replaces it with Torch without CUDA version 2.1.2
- Running StoryToolkitAI 0.24 confirms that the CUDA option is not available.
**Installation log file
Installation Log.txt
Expected behavior
I expect Torch with CUDA should not be replaced with Torch without CUDA.
System (please complete the following information):
- Windows 11
- Python versions 3.10.0rc2
- StoryToolkitAI Version 0.24
Additional context
Anything else that is relevant here.
Hey, thanks for pointing this out!
This is most likely because the tool does a pip install -r requirements.txt on that particular update.
We'll add a step which will check if there's CUDA support or not and let it decide which version to keep/install.
Cheers!
Thanks