Stuck getting 'Got an invalid category_id=1 for a dataset of 1 classes' error despite cfg

[paleheart]

Tried this originally on windows (which I know is a no, no) So I retried under an ubuntu instance on python 3.8 and 3.9

My dataset function seems to be ok and I got all the way through training with a single class but after wiping out my output folder and adding a second class to retry training I keep getting this error

AssertionError: Got an invalid category_id=1 for a dataset of 1 classes

I set the cfg to:

cfg.MODEL.ROI_HEADS.NUM_CLASSES = 2

(also tried 3 in case it counted one for ground)

which seems to be the right one for the model I was trying to load (mask_rcnn_R_101_FPN_3x)

ive tried it without everything but creating the cfg and setting the number of classes but I still get the error saying my dataset only has one class as soon as I call

trainer = DefaultTrainer(cfg)

I googled around and tried stackoverflow but havent gotten any responses yet and havent seen a lot of people talking about this

is there somewhere besides the cfg I need to set the number of classes?

[paleheart]

I figured it out

I was doing

MetadataCatalog.get("my_dataset").(thing_classes=thing_list

when I should have been doing

MetadataCatalog.get("my_dataset").set(thing_classes=thing_list)