# 覺得哪裡怪怪的請趕快跟助教說

for 多分類：

1. "output" layer: units=類別數, activation='softmax'

2. loss = tf.keras.losses.SparseCategoricalCrossentropy()) # y沒有one hot的情況

3. prediction = np.argmax(model.predict(你設定的input list), axis=1)

4. validation: acc, recall, precision, ...