Error rate gets worse after unfreezing

Hi everyone, I’ve stumbled onto the Stanford Dogs dataset and I wanted to try to build an image classification model for it.
After the first training of the model I got an error_rate of around 0.235, to improve my error_rate I plotted the learning rate (taking Jeremys advice on how to pick a good learning rate for 2nd round of training) and finally unfroze and trained my CNN again.
But this time, my error rate was actually worse and I even overfitted.
I’m lost at this point, I’ve been trying different learning rates for the past couple of hours but I can’t seem to find my error.
If somebody could have a look at my code and point out my error, I would really appreciate that.