如何從兩個順序網絡創建自動編碼器？

Question

我有兩個順序網絡（編碼器網絡和解碼器網絡）。 如何使用順序 API 創建自動編碼器 model？

請不要推薦使用功能 API 或解釋功能優於順序的好處，因為這不是這里的問題。

encoder_network = tf.keras.Sequential([
    Conv2D(64, 3, padding='same', activation="swish"),
    DownscaleBlock(1),
    DownscaleBlock(2),
    Conv2D(128, 3, padding='same', activation="swish"),
    Conv2D(32, 3, padding='same', activation="swish"),
    Conv2D(10, 3, padding='same'),
])

decoder_network = tf.keras.Sequential([
    Conv2D(4, 3, padding='same', activation="swish"),
    Conv2D(16, 3, padding='same', activation="swish"),
    Conv2D(64, 3, padding='same', activation="swish"),
    UpscaleBlock(1),
    UpscaleBlock(2),
    Conv2D(4, 3, padding='same', activation="swish"),
    Conv2D(1, 3, padding='same'),
])

Answer 1

您可以像使用圖層一樣使用模型：

model = tf.keras.Sequential([
    encoder_network,
    decoder_network
])