尝试存储神经网络时获取“无法腌制 _thread.RLock 对象”

Question

I am currently training a neural network and I try to store the trained model for future use. The model is based on Sequential from keras (see below). I am using joblib.dump(model, output_file_gen) to store the information. However, I get the error message:

TypeError: can't pickle _thread.RLock objects.

I have looked at some StackOverflow posts regarding this error message and it seems to relate to multithreading. I am not sure what happens in the model but maybe somebody can give me advice how to store the model either by taking steps to get rid of this error or by suggesting a better route to store a neural network.

NN setup is included below:

model = Sequential()

model.add(Dense(256, input_dim=self.latent_dim))
model.add(LeakyReLU(alpha=0.2))
model.add(BatchNormalization(momentum=0.8))
model.add(Dense(512))
model.add(LeakyReLU(alpha=0.2))
model.add(BatchNormalization(momentum=0.8))
model.add(Dense(1024))
model.add(LeakyReLU(alpha=0.2))
model.add(BatchNormalization(momentum=0.8))
model.add(Dense(np.prod(self.img_shape), activation='tanh'))
model.add(Reshape(self.img_shape))

Answer 1

It is not recommended to use pickle or cPickle to save a Keras model.which is the cause of the error here (loosely reasoned)

You can use model.save(filepath) to save the model into a single HDF5 file which will contain:

1. the architecture of the model, allowing to re-create the model
2. the weights of the model
3. the training configuration (loss, optimizer)
4. the state of the optimizer, allowing to resume training exactly where you left off.

You can then use keras.models.load_model(filepath) to reinstantiate/reload your model.

The above will use a lot of disk space. so you can alternatively save the model weights. see here for more details