前言:
經過多次的測試和訓練
val_accuracy 在最後幾乎都是處於0.6500左右的狀態
所以這次我想從optimizer著手
也許是收斂過早的問題

net_final.compile(optimizer=Adam(learning_rate=0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-08),
                  loss='categorical_crossentropy', 
                  metrics=['accuracy'])

104/104 [==============================] - 128s 1s/step - batch: 51.5000 - size: 63.6154 - loss: 4.1122 - accuracy: 0.1167 - val_loss: 3.2964 - val_accuracy: 0.2344
儲存訓練模型
104/104 [==============================] - 130s 1s/step - batch: 51.5000 - size: 63.6154 - loss: 3.9934 - accuracy: 0.1313 - val_loss: 3.1710 - val_accuracy: 0.2524
儲存訓練模型
104/104 [==============================] - 136s 1s/step - batch: 51.5000 - size: 63.6154 - loss: 3.8718 - accuracy: 0.1471 - val_loss: 3.0487 - val_accuracy: 0.2608
訓練中...