run_model.py

#Train model and get metrics. 
model.compile(
    optimizer='adam',
    loss='binary_crossentropy',
    metrics=['accuracy'],    
)

from keras.callbacks import Callback

class Histories(Callback):
    
    def on_train_begin(self,logs={}):
        self.batch_loss = []
        self.batch_accuracy = []
        self.loss = []
        self.accuracy = []
        self.val_loss = []
        self.val_accuracy = []
    
    def on_train_batch_end(self, batch, logs={}):
        self.batch_loss.append(logs.get('loss',-1))
        self.batch_accuracy.append(logs.get('accuracy',-1))               
        
    def on_epoch_end(self, epoch, logs={}):
        self.loss.append(logs.get('loss',-1))
        self.accuracy.append(logs.get('accuracy',-1))
        self.val_loss.append(logs.get('val_loss',-1))
        self.val_accuracy.append(logs.get('val_accuracy',-1))

h_cb = Histories()

#Fit model in 3 epochs.  Save metrics to history.
history = model.fit(train_ds, validation_data=val_ds, epochs=3, callbacks=[h_cb])