python – why confusion matrix is not compatible with the accuracy?

I am doing Bert fine tuning on my multi class text classification.
The accuracy and Loss plots are as below.

1 - my learning rate for Adam is learning_rate= 2e-5, which learning rate is better to start that I my accuracy start to train from low number rather that 79%, and it goes up after some echos.I mean for example it starts from 50% to 70% and it goes up.

2- why my confusion matrix shows that I have a very few accuracy?

enter image description here

enter image description here

enter image description here

Product of the Month September 2016

Source link

Leave a Comment

Your email address will not be published. Required fields are marked *