Hi,
I am trying to display the model confidence for each class label for the MultiCategory learner. To achieve this, I am using the probs
from the get_preds()
method and then running the following code:
preds, probs = learn.get_preds()
confs = torch.mean(probs.sigmoid(), 0)
confs = {l:round(c.item(),4) for l,c in zip(dls.vocab, confs)}
confs = sorted(confs.items(), key=operator.itemgetter(1),reverse=True)
confs
Is this the right approach and is there a better way?
Any comments will be much appreciated.
Thanks in advance,
Best Regards,
Bilal