Since you are using functional API for creating the autoencoder, the best way to reconstruct the encoder and decoder is to use the functional API and the Model class again: autoencoder= K.models.load_model('fashion-autoencoder.hdf5')Įncoder = Model(autoencoder.input, )ĭecoder_input = Input(shape=(encoding_dim,))ĭecoder = Model(decoder_input, autoencoder.layers(decoder_input)) So, my general question is how to extract parts of loaded models. I even have less success extracting the decoder (from the saved autoencoder) since I cannot use push() and tried stuff like decoder = decoder.layers but it did not work. So, my extraction for the encoder did not work since the dimensions are not correct. Then I load images again (not shown) and use the encoder encoded_imgs = encoder.predict(x_test) That I understand in a kind of way but do not know how important it is. 'Discrepancy between trainable weights and collected trainable'
# delete the last layers to get the encoderĪnd the encoder looks the same as the original in step one what makes me think the extraction has worked well: Layer (type) Output Shape Param #īut I also get the warning training.py:478: UserWarning: Discrepancy between trainable weights and collected trainable weights, did you set `ainable` without calling `pile` after ? I load the model using encoder= K.models.load_model('fashion-autoencoder.hdf5') So lets go to step 2 where I have my problems. Later, I load the images (not shown) and do the predictions like # encode and decode some images from test setĭecoded_imgs = decoder.predict(encoded_imgs) In my real example I save it with a callback so a workaround by saving the encoder and decoder does not seem a real solution. So I train the model and save it by autoencoder.save('fashion-autoencoder.hdf5'). # encoder: map an input to its encoded representationĮncoded_input = Input(shape=(encoding_dim,))ĭecoder = Model(encoded_input, decoder_layer(encoded_input)) # full AE model: map an input to its reconstruction
I have problems (see second step) to extract the encoder and decoder layers from the trained and saved autoencoder.įor step one I have the very simple network as follows: input_img = Input(shape=(784,))Įncoded = Dense(encoding_dim, activation='relu')(input_img)ĭecoded = Dense(784, activation='sigmoid')(encoded) Use this best model (manually selected by filename) and plot original image, the encoded representation made by the encoder of the autoencoder and the prediction using the decoder of the autoencoder.That process can be some weeks before the following part. Load the images, do the fitting that may take some hours or days and use a callback to save the best autoencoder model.I want to divide the autoencoder learning and applying into two parts following and using the fashion-mnist data for testing purposes:
0 kommentar(er)
