hi,
What is the challenge policy of choosing model sanphot 
Because we are training only on 100 samples, we are probably overfitting
Do you take the model from the last epoch or the best model? 
Best
Eli