How to fix ("ValueError: You are trying to load a weight file containing 16 layers into a model with 0 layers")

Question

我是这样使用 vgg16 的：

model = VGG16()
data, labels = ReadImages(TRAIN_DIR)

vgg16 = VGG16()

model = Sequential()

#Converting VGG16 into Sequential model
for layer in vgg16.layers[:-1]:
    model.add(layer)

#Freezing all layers except last layer for transfer learning
for layer in model.layers:
    layer.trainable = False

#Adding custom softmax layer
model.add(Dense(1,activation='sigmoid'))

#Compiling our model
model.compile(optimizer='adam',loss='binary_crossentropy',metrics=['accuracy'])

model.fit(np.array(data), np.array(labels), batch_size=32, epochs=3)

model.save('model.h5')

当我试图在另一个 py 文件中加载这个模型时..:[=13=]

model = load_model('model.h5')

我已经尝试过 load_weights 并且也抛出错误

...returns 这个错误：

ValueError: You are trying to load a weight file containing 16 layers into a model with 0 layers

我应该怎么做才能加载此模型以进行预测？

版本：keras 2.2.4 tensorflow 1.14.0

Answer 1

已知问题：https://github.com/keras-team/keras/issues/10417

共有三个选项：1. 重新创建模型架构并使用“load_weights”。如果您只想进行预测，这很好。 2. 降级到 Keras 版本 2.1.6. 3. 在此 link https://github.com/keras-team/keras/issues/10417#issuecomment-435620108 上可用的解决方法。我为 VGG16 改编了这个。这会更新 h5 文件。

def fix_layer0(filename, batch_input_shape, dtype):
    with h5py.File(filename, 'r+') as f:
        model_config = json.loads(f.attrs['model_config'].decode('utf-8'))
        layer0 = model_config['config']['layers'][0]['config']
        layer0['batch_input_shape'] = batch_input_shape
        layer0['dtype'] = dtype
        f.attrs['model_config'] = json.dumps(model_config).encode('utf-8')

fix_layer0('model.h5', [None, 224, 224, 3], 'float32')

loaded_model = load_model('model.h5')

Answer 2

此问题已在较新版本的 TensorFlow 中得到解决

只需升级您的 TensorFlow： pip 安装 tensorflow --upgrade

How to fix ("ValueError: You are trying to load a weight file containing 16 layers into a model with 0 layers")

How to fix ("ValueError: You are trying to load a weight file containing 16 layers into a model with 0 layers")

python

machine-learning

image-processing

keras

vgg-net