Moritz Nottebaum comments

Repositories
Issues
Comments

Results 4 comments of


                                            Moritz Nottebaum

ModuleNotFoundError: No module named 'mmdet'

You can train your SAM model without mmdet/mmcv (I did it). Just put the imports into the if-conditions and not at the beginning of the document. The standard configuration doesn't...

AssertionError assert (boxes1[:, 2:] >= boxes1[:, :2]).all()

It's definetely not only when you train on a custom dataset. Basically if you change anything in the model, the error can occur frequently.

AssertionError assert (boxes1[:, 2:] >= boxes1[:, :2]).all()

You can also deactivate the amp (automatic mixed precision...) during training. If that's not possible, as training'd need too much memory, you can deactivate it just for the attention operations:...

AssertionError assert (boxes1[:, 2:] >= boxes1[:, :2]).all()

> @csampat-a, that's interesting. Yes, I was able to run multiple full training runs also on a custom dataset with my suggested fix. Before, I ran into the exact same...