tensorflow-gpu ValueError:未知层:功能

Question

我正在尝试从一个(由mobilenetv2训练的) .h5文件加载一个模型。

from tensorflow.keras import models
class Classifier(object):

    def __init__(self, model_path):
        self.model = models.load_model(model_path)

我的模型代码：

model_pre = keras.applications.MobileNetV2(
    input_shape=(224, 224, 3),
    include_top=False,
)

model_pre.trainable = False

last_output = model_pre.output

x = Flatten()(last_output)
x = Dense(3, activation='softmax')(x) 
model = tf.keras.Model(model_pre.input, x)

model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])

model.summary()

tensorflow-gpu == 2.2 (具有2.3相同问题) python == 3.8.3

加载后，我得到了这个问题

2020-10-12 21:37:39.500606: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudart64_101.dll
Traceback (most recent call last):
  File "C:/Users/manos/projects/gesture_recognition/app.py", line 28, in <module>
    classifier = Classifier(CLASSIFIER_MODEL_PATH)
  File "C:\Users\manos\projects\gesture_recognition\classifier\gesture_classifier.py", line 9, in __init__
    self.model = models.load_model(model_path)
  File "C:\Users\manos\AppData\Roaming\Python\Python38\site-packages\tensorflow\python\keras\saving\save.py", line 184, in load_model
    return hdf5_format.load_model_from_hdf5(filepath, custom_objects, compile)
  File "C:\Users\manos\AppData\Roaming\Python\Python38\site-packages\tensorflow\python\keras\saving\hdf5_format.py", line 177, in load_model_from_hdf5
    model = model_config_lib.model_from_config(model_config,
  File "C:\Users\manos\AppData\Roaming\Python\Python38\site-packages\tensorflow\python\keras\saving\model_config.py", line 55, in model_from_config
    return deserialize(config, custom_objects=custom_objects)
  File "C:\Users\manos\AppData\Roaming\Python\Python38\site-packages\tensorflow\python\keras\layers\serialization.py", line 105, in deserialize
    return deserialize_keras_object(
  File "C:\Users\manos\AppData\Roaming\Python\Python38\site-packages\tensorflow\python\keras\utils\generic_utils.py", line 361, in deserialize_keras_object
    (cls, cls_config) = class_and_config_for_serialized_keras_object(
  File "C:\Users\manos\AppData\Roaming\Python\Python38\site-packages\tensorflow\python\keras\utils\generic_utils.py", line 321, in class_and_config_for_serialized_keras_object
    raise ValueError('Unknown ' + printable_module_name + ': ' + class_name)
ValueError: Unknown layer: Functional

我到处寻找这个问题的解决方案，他们写了关于升级或降级tensorflow-gpu版本的文章，但没有一个选项对我没有帮助

Answer 1

当我使用import：from tensorflow.keras.callbacks import ModelCheckpoint保存模型，但只使用import from keras.models import load_model加载保存的模型时，我遇到了这个问题。一旦我把它改成了from tensorflow.keras.models import load_model，它就工作得很好。