Prevent Comfyui Custom Nodes From Overwriting Pytorch Rocm With CUDA Version

This project addresses the issue of ComfyUI custom nodes inadvertently overwriting the PyTorch ROCm setup with the CUDA version, which could lead to compatibility problems in machine learning workflows. It ensures that the ROCm version of PyTorch remains intact while allowing users to utilize ComfyUI's features without interference.

The project leverages Python scripting and environment management tools to create a safeguard that maintains the integrity of the ROCm installation. It is particularly useful for developers and researchers working in environments where AMD GPUs and ROCm are prevalent, ensuring that their machine learning models can execute properly without the risk of unintentional modifications to the underlying framework.