Closed qianyue76 closed 3 months ago
I get the same error when I do the inference step. Could you please provide more details about the issue you have done, thank you very much!
I add the comited code to the nms methods, but still encounter this error:'tuple' object has no attribute 'shape'
This issus was solved by a little change:
if isinstance(prediction, dict) and 'one2one' in prediction: # model.pt, output = {"one2many": torch.Tensor, "one2one": torch.Tensor }
prediction = prediction['one2one'][0]
A small fix with
non_max_suppression
inops.py
. The fix originated from the ultralytics issue 13587 issues:13587