Can someone explain me why is there a difference in error rates in this case? When I use fine_tine(8), the error rate is 0.05 after 7th epoch. If I run fine_tune(7), I would expect the 7th epoch to have the same error. But that isn’t the case.
It is because of randomized seeds. When you wanted to fine tune until the 8th epoch, the seed was different as compared to the seed when you fine tuned it till the 7th epoch. For more information about this, read Role of Randomness in Machine Learning. You don’t need to go into the nitty-gritty details of this topic, but this will be very helpful if you want to understand more about what I just said.