Does the graph below from the c2w3 assignment look alright?
I am reaching 99.9% accuracy within 7 epochs which sounds too good to be true.
Might I be doing something wrong?
Hope this text from the notebook helps: The training should have stopped after less than 10 epochs
Hello @ankitprashnani I am not TF1 mentor but from the look, plots are diverging too much
Please refer to this video to Understand about the training and validation loss curves Thank you
But I found the number of epochs could vary. I found it converging after exactly 10, so I ran it again, which made it conclude just after 1 epoch.
Training from scratch should achieve the threshold within 10 epochs. Your method, though technically valid, is not the right way to go about solving the assignment. Good luck with creating a better architecture.