Now have ability to pass loss functions and associated hyperparameters as CLI. Should be very transparent to user with all the dirty work done in loss_functions.py.
Should be very transparent to user: the loss function CLI is a string which then gets converted into a function in loss_functions.py. A dictionary maps a given string to (1) a function and (2) a list of associated hyperparameters. The function choose_loss then converts this into an acceptable format to pass to model.compile.
Now have ability to pass loss functions and associated hyperparameters as CLI. Should be very transparent to user with all the dirty work done in
loss_functions.py
.Should be very transparent to user: the loss function CLI is a string which then gets converted into a function in
loss_functions.py
. A dictionary maps a given string to (1) a function and (2) a list of associated hyperparameters. The functionchoose_loss
then converts this into an acceptable format to pass tomodel.compile
.