Prabhat Singh

> No, when building custom cuda extension local cuda toolkit should match to the pytorch's cuda support version , hence you need to install cuda 11.8 version https://developer.nvidia.com/cuda-11-8-0-download-archive

Not sure if you can several cuda versions or not, but yess the cuda version in local must match your pytorch version only for the cases where you build **custom...

 Here's another sample when I detect dog , directly through the notebook

Next I tried to load the model as mentioned in the [notebook](https://colab.research.google.com/github/roboflow-ai/notebooks/blob/main/notebooks/zero-shot-object-detection-with-grounding-dino.ipynb#scrollTo=rH62Frfih5lY) , and here the results were good for example the combined prompt results were as shown ``` TEXT_PROMPT...

No, I tested it by using `,` as a separator only.