Feature/loss tracker for dropout

[beckstev]

Loss tracker for dropout

Added the functionality to save the loss and acc after each epoch whenever we are using PreDogNN or PreBigDogNN. Therefore, I modified your HistoryEpoch class. In addition, I also merged the plot files for loss and acc into a single file containing two subplots.

How to test

python training.py <PreDogNN or PreBigDogNN> <#races associated encoder> -n <associated #races> -e 1

---

Hypertuning

• Adapted the z-axis of the 3d scatter plot
• Removed bs 17 for hyper tunning, because it tensorflow always failed

[beckstev]

Unfortunately, this raises an error. ValueError: Length of values does not match length of index, while saving the learing-rate in the function save_history_epoch. It appears if i train 3 Epochs instead of 1, when i train only one epoch everthing is fine. Can you reproduce this error?

Addressed with the latest commit!