I’m using these lines to load data from .npy files:

X_train = np.load(r'drive/MyDrive/X_train.npy', mmap_mode='r') X_test = np.load(r'drive/MyDrive/X_test.npy', mmap_mode='r') y_test = np.load(r'drive/MyDrive/y_test.npy', mmap_mode='r') y_train = np.load(r'drive/MyDrive/y_train.npy', mmap_mode='r')

Then use I fit directly the X_train and y_train into tf.keras.model object:

model.fit(X_train, y_train, ...)

Is this the recommended way to do it? It seems to work! Do I need to create a data generator or something using this method tf.data.Dataset.from_generator?