fnn_helper.py

import keras
import matplotlib.pyplot as plt
from IPython.display import clear_output

class PlotLosses(keras.callbacks.Callback):
    def __init__(self, plot_interval=1, evaluate_interval=10, x_val=None, y_val_categorical=None):
        self.plot_interval = plot_interval
        self.evaluate_interval = evaluate_interval
        self.x_val = x_val
        self.y_val_categorical = y_val_categorical
        #self.model = model
    
    def on_train_begin(self, logs={}):
        print('Begin training')
        self.i = 0
        self.x = []
        self.losses = []
        self.val_losses = []
        self.acc = []
        self.val_acc = []
        self.logs = []
        self.weights = []
        self.acc_2 = []
        self.loss_2 = []
    
    def on_epoch_end(self, epoch, logs={}):
        if self.evaluate_interval is None:
            self.logs.append(logs)
            self.x.append(self.i)
            self.losses.append(logs.get('loss'))
            self.val_losses.append(logs.get('val_loss'))
            self.acc.append(logs.get('acc'))
            self.val_acc.append(logs.get('val_acc'))
            self.weights.append(self.model.get_weights())
            self.i += 1
        
        if (epoch%self.plot_interval==0):
            clear_output(wait=True)
            f, (ax1, ax2) = plt.subplots(1, 2, sharex=True, figsize=(20,5))
            ax1.plot(self.x, self.losses, label="loss")
            ax1.plot(self.x, self.val_losses, label="val_loss")
            ax1.legend()

            ax2.plot(self.x, self.acc, label="acc")
            ax2.plot(self.x, self.val_acc, label="val_acc")
            ax2.legend()
            plt.show();
        #score = self.model.evaluate(x_test, y_test_categorical, verbose=0)
        
        #print("accuracy: ", score[1])
    
    def on_batch_end(self, batch, logs={}):
        if self.evaluate_interval is not None:
            if (batch%self.evaluate_interval==0):
                self.i += 1
                self.logs.append(logs)
                self.x.append(self.i)
                self.losses.append(logs.get('loss'))
                self.acc.append(logs.get('acc'))

                if self.x_val is not None:
                    score = self.model.evaluate(self.x_val, self.y_val_categorical, verbose=0)
                    self.val_losses.append(score[0])
                    self.val_acc.append(score[1])