Model classification metrics are much more significant than training and testing losses.
def visualize_training_logs(test_losses, training_losses):
epochs = np.arange(1, len(test_losses) + 1)
plt.plot(epochs, test_losses, label="test losses", color="r")
plt.plot(epochs, training_losses, label="train losses", color="gray")
plt.xlabel("epochs")
plt.ylabel("loss")
plt.title("training logs visualization")
plt.legend()
plt.show()
we can see strong overfitting in the graph below. However the model seems to learn something at the beginning.
