Fix possible pytorch exception

[StevePotter]

Describe your changes and approach used

I did not file an issue, and instead just made a PR when I encountered the problem.

I am using this great model in conjunction with a few other pytorch models. I ran LangSAM first, then the other model. That worked fine. But when I run LangSAM, it failed with the following stack trace:

    masks, _, phrases, _ = self.langSam.predict(rbg_image, segment_prompt)
  File "/opt/conda/envs/my/lib/python3.8/site-packages/lang_sam/lang_sam.py", line 119, in predict
    boxes, logits, phrases = self.predict_dino(image_pil, text_prompt, box_threshold, text_threshold)
  File "/opt/conda/envs/my/lib/python3.8/site-packages/lang_sam/lang_sam.py", line 102, in predict_dino
    boxes = box_ops.box_cxcywh_to_xyxy(boxes) * torch.Tensor([W, H, W, H])
RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!

I did some investigation and the problem comes from the boxes from the GroundingDINO predict function are CPU but the torch.Tensor was using cuda. So the fix was simply to use the device from boxes in the torch.Tensor call.

I tested and it works fine. Now I can use multiple models in the same process without trouble. Thanks!

Checklist before requesting a review

• [x] I have performed a self-review of my code.
• [x] I have performed linting on my code.
• [x] N/A I have linked the related issue.
• [x] The code is documented accordingly.
• [x] If it is a core feature, I have added thorough tests.