包含 EfficientNet 子模型的 Keras 模型无法加载权重； VGG 工作正常

Question

我有一个超级模型，它使用子模型作为特征提取层。我的代码是模块化的，因此我可以轻松地通过更改我分配的任何子模型来切换我正在使用的子模型来执行特征提取：

...

elif FEATURE_EXTRACTOR == "VGG16":

    Features = keras.applications.VGG16(

        weights = "imagenet",
        pooling = FEATURE_POOLING,
        include_top = False

    )
elif FEATURE_EXTRACTOR == "EfficientNetB0":

    Features = keras_applications_master.keras_applications.efficientnet.EfficientNetB0(
    # ^ Local copy of official keras repo: https://github.com/keras-team/keras-applications
    # because pip install --upgrade keras doesn't install version with efficientnet.

        weights = "imagenet",
        include_top = False,
        pooling = FEATURE_POOLING,
        classes = None

    )

...

我保存和加载 ubermodel 及其权重的例程也知道哪个子模型用于特征提取：

model.load_weights(submodel_specific_path)

对于任何子模型，我都可以执行初始训练运行并将 ubermodel 保存到磁盘。如果我尝试继续训练或微调任何包含 VGG16 子模型的 ubermodel，使用 load_weights 加载权重，一切正常。但是当我load_weights 任何具有高效网络子模型（或者说，keras.applications.xception.Xception）的 ubermodel 时，我收到以下错误：

Traceback (most recent call last):
  File "image_model.py", line 284, in <module>
    model.load_weights(model_path)
  File "C:\Users\Username\Anaconda3\envs\tensorflow1\lib\site-packages\keras\engine\saving.py", line 492, in load_wrapper
    return load_function(*args, **kwargs)
  File "C:\Users\Username\Anaconda3\envs\tensorflow1\lib\site-packages\keras\engine\network.py", line 1227, in load_weights
    reshape=reshape)
  File "C:\Users\Username\Anaconda3\envs\tensorflow1\lib\site-packages\keras\engine\saving.py", line 1294, in load_weights_from_hdf5_group_by_name
    reshape=reshape)
  File "C:\Users\Username\Anaconda3\envs\tensorflow1\lib\site-packages\keras\engine\saving.py", line 861, in preprocess_weights_for_loading
    weights = convert_nested_model(weights)
  File "C:\Users\Username\Anaconda3\envs\tensorflow1\lib\site-packages\keras\engine\saving.py", line 836, in convert_nested_model
    original_backend=original_backend))
  File "C:\Users\Username\Anaconda3\envs\tensorflow1\lib\site-packages\keras\engine\saving.py", line 980, in preprocess_weights_for_loading
    weights[0] = np.transpose(weights[0], (3, 2, 0, 1))
  File "<__array_function__ internals>", line 6, in transpose
  File "C:\Users\Username\Anaconda3\envs\tensorflow1\lib\site-packages\numpy\core\fromnumeric.py", line 651, in transpose
    return _wrapfunc(a, 'transpose', axes)
  File "C:\Users\Username\Anaconda3\envs\tensorflow1\lib\site-packages\numpy\core\fromnumeric.py", line 61, in _wrapfunc
    return bound(*args, **kwds)
ValueError: axes don't match array

我做错了什么？

Answer 1

我无法重现您的问题，但我发现了这个Github Issue，它说当我们使用多 GPU 模型时会出现这个问题。所以诀窍应该是：

multi_gpu_model = multi_gpu_model(model, gpus=G)
model.save_model(...)

代替

multi_gpu_model = multi_gpu_model(model, gpus=G)
multi_gpu_model.save_model(...)

话虽如此，我能够成功地创建EfficientNetB0 模型，评估模型，保存模型并最终加载模型。

使用 model.save 构建、评估和保存模型的代码 -

%tensorflow_version 1.x
import keras
from keras.models import Model
from keras.layers import GlobalAveragePooling2D, Dense, Flatten
from keras_efficientnets import EfficientNetB0

model = EfficientNetB0(input_shape=(224, 224, 3), classes=1000, include_top=False, weights='imagenet')

x = model.output
x = GlobalAveragePooling2D()(x)
x = Dense(20, activation='relu')(x)
x = Dense(17, activation='softmax')(x)
model = Model(inputs = model.input, outputs = x)

# summarize model
#model.summary()

# (2) Get Data
import tflearn.datasets.oxflower17 as oxflower17
x, y = oxflower17.load_data(one_hot=True)

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

model.fit(x, y, batch_size=64, epochs= 1, verbose=1, validation_split=0.2, shuffle=True)

# evaluate the model
scores = model.evaluate(x, y, verbose=0)
print("%s: %.2f%%" % (model.metrics_names[1], scores[1]*100))

# save model and architecture to single file
model.save("model.h5py")
print("Saved model to disk")

输出 -

TensorFlow 1.x selected.
Using TensorFlow backend.
WARNING:tensorflow:From /tensorflow-1.15.2/python3.6/tensorflow_core/python/ops/resource_variable_ops.py:1630: calling BaseResourceVariable.__init__ (from tensorflow.python.ops.resource_variable_ops) with constraint is deprecated and will be removed in a future version.
Instructions for updating:
If using Keras pass *_constraint arguments to layers.
WARNING:tensorflow:From /tensorflow-1.15.2/python3.6/tflearn/helpers/summarizer.py:9: The name tf.summary.merge is deprecated. Please use tf.compat.v1.summary.merge instead.

WARNING:tensorflow:From /tensorflow-1.15.2/python3.6/tflearn/helpers/trainer.py:25: The name tf.summary.FileWriter is deprecated. Please use tf.compat.v1.summary.FileWriter instead.

WARNING:tensorflow:From /tensorflow-1.15.2/python3.6/tflearn/collections.py:13: The name tf.GraphKeys is deprecated. Please use tf.compat.v1.GraphKeys instead.

WARNING:tensorflow:From /tensorflow-1.15.2/python3.6/tflearn/config.py:123: The name tf.get_collection is deprecated. Please use tf.compat.v1.get_collection instead.

WARNING:tensorflow:From /tensorflow-1.15.2/python3.6/tflearn/config.py:129: The name tf.add_to_collection is deprecated. Please use tf.compat.v1.add_to_collection instead.

WARNING:tensorflow:From /tensorflow-1.15.2/python3.6/tflearn/config.py:131: The name tf.assign is deprecated. Please use tf.compat.v1.assign instead.

Downloading Oxford 17 category Flower Dataset, Please wait...
100.0% 60276736 / 60270631
('Succesfully downloaded', '17flowers.tgz', 60270631, 'bytes.')
File Extracted
Starting to parse images...
Parsing Done!
WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:422: The name tf.global_variables is deprecated. Please use tf.compat.v1.global_variables instead.

WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:431: The name tf.is_variable_initialized is deprecated. Please use tf.compat.v1.is_variable_initialized instead.

WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:438: The name tf.variables_initializer is deprecated. Please use tf.compat.v1.variables_initializer instead.

Train on 1088 samples, validate on 272 samples
Epoch 1/1
1088/1088 [==============================] - 203s 187ms/step - loss: 1.6433 - accuracy: 0.5561 - val_loss: 1.9315 - val_accuracy: 0.5074
accuracy: 54.85%

使用 load_model 加载模型并评估 -

%tensorflow_version 1.x
# load and evaluate a saved model
from numpy import loadtxt
from keras.models import load_model

# load model
model = load_model('model.h5py')

# summarize model
#model.summary()

# (2) Get Data
import tflearn.datasets.oxflower17 as oxflower17
x, y = oxflower17.load_data(one_hot=True)

# evaluate the model
score = model.evaluate(x, y, verbose=0)
print("%s: %.2f%%" % (model.metrics_names[1], score[1]*100))

输出 - 模型在保存前和加载后的准确度应该匹配，在这里也是如此。

accuracy: 54.85%

使用 model.save_weights 保存模型并使用 model.load_weights 加载模型 -

from keras.models import model_from_json

# serialize model to JSON
model_json = model.to_json()
with open("model.json", "w") as json_file:
    json_file.write(model_json)
# serialize weights to HDF5
model.save_weights("model.h5")
print("Saved model to disk")

# later...

# load json and create model
json_file = open('model.json', 'r')
loaded_model_json = json_file.read()
json_file.close()
loaded_model = model_from_json(loaded_model_json)
# load weights into new model
loaded_model.load_weights("model.h5")
print("Loaded model from disk")

# evaluate loaded model on test data
loaded_model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
score = loaded_model.evaluate(x, y, verbose=0)
print("%s: %.2f%%" % (loaded_model.metrics_names[1], score[1]*100))

输出 -

Saved model to disk
Loaded model from disk
accuracy: 54.85%

希望这能回答您的问题。快乐学习。