Accuracy less than 100, but no mistakes

Hello all, I am training a model using Huggingface and when evaluating it I get approx 80% accuracy.
I then try to plot the confusion matrix (using pycm) but I see no error.
Any ideas what this might be?

Could you be providing the same labels to the confusion matrix plotting function, like plotting predictions versus predictions or ground truth versus ground truth? That kind of mistake is one that I make all the time.