Mask2FormerForUniversalSegmentation object argument after ** must be a mapping, not Tensor

#1
by yyyyy611 - opened

Hi there! I got a error said 'Mask2FormerForUniversalSegmentation object argument after ** must be a mapping, not Tensor'. It seems that the type of video_input is incorrect. Do you know how to fix it? Thanks!

Hey the PR for video mask2former is not merged yet AFAIK.
I was working on this a while back but then got busy with some other work.

I'll try to get back to this soon and wrap it up.

OK. Thank you for your work!

Hi!

I am trying to reproduce the code snippet but I face 3 errors:

  1. The model is not under "facebook/" but under "shivalikasingh/"
  2. The AutoImageProcessor does not recognise the model. I changed it to Mask2FormerImageProcessor as workaround. But then,
  3. the processor does not have the method post_process_video_instance_segmentation

Could you guide me on how to use your model please? I would like to fine-tune it for panoptic segmentation.

Thanks!
Ramon

Sign up or log in to comment