CUDA environment setup issues with RTX 5070 Ti (dependencies mismatch)

[ga3ll3y]

Hello,

I'm currently facing issues running a demo due to CUDA environment and dependency mismatches. I'm working with an RTX 5070 Ti, and it's been difficult to find compatible versions of PyTorch, CUDA, and related libraries that work properly together.

So far, I’ve encountered problems such as:

• Incompatible CUDA / PyTorch versions
• Missing or conflicting dependencies
• Some operators not available on CUDA
• Errors related to NumPy / OpenCV compatibility

I would really appreciate any guidance on:

• A stable environment setup (CUDA + PyTorch versions) for this GPU
• Recommended installation steps or working configurations
• Any known issues with newer GPUs and current deep learning frameworks

Thanks a lot for your help!