ValueError: Unable to load weights 03_04_vae_digits_analysis.ipynb

[Lana2548]

From Japan.
I'm not good at English, so it may be hard to read. Sorry.

• Anaconda
• Nvidia GTX 1070

When I used load_model() in 03_04_vae_digits_analysis.ipynb, I got an error message.

---

ValueError Traceback (most recent call last)
in
----> 1 vae = load_model(VariationalAutoencoder, RUN_FOLDER)

~\Desktop\study\GDL_code\utils\loaders.py in load_model(model_class, folder)
138 model = model_class(*params)
139
--> 140 model.load_weights(os.path.join(folder, 'weights/weights.h5'))
141
142 return model

~\Desktop\study\GDL_code\models\VAE.py in load_weights(self, filepath)
204
205 def load_weights(self, filepath):
--> 206 self.model.load_weights(filepath)
207
208 def train(self

~\AppData\Roaming\Python\Python37\site-packages\tensorflow\python\keras\engine\training.py in load_weights(self, filepath, by_name, skip_mismatch, options)
2221 if not self._is_graph_network and not self.built:
2222 raise ValueError(
-> 2223 'Unable to load weights saved in HDF5 format into a subclassed '
2224 'Model which has not created its variables yet. Call the Model '
2225 'first, then load the weights.')

ValueError: Unable to load weights saved in HDF5 format into a subclassed Model which has not created its variables yet. Call the Model first, then load the weights.

Please give me any advice.
Thank you.

[eiler-partner]

Hi,
I am using the code from tensorflow_2 directory
I have the same problem:

---

ValueError Traceback (most recent call last)
in
----> 1 vae = load_model(VariationalAutoencoder, RUN_FOLDER)

~\Deep_Learning\H_Generative_Deep_Learning\utils\loaders.py in load_model(model_class, run_folder)
138 model = model_class(*params)
139
--> 140 model.load_weights(os.path.join(run_folder, 'weights/weights.h5'))
141
142 return model

~\Deep_Learning\H_Generative_Deep_Learning\models\VAE.py in load_weights(self, filepath)
204
205 def load_weights(self, filepath):
--> 206 self.model.load_weights(filepath)
207
208 def train(self, x_train, batch_size, epochs, run_folder, print_every_n_batches = 100, initial_epoch = 0, lr_decay = 1):

~\anaconda3\lib\site-packages\tensorflow\python\keras\engine\training.py in load_weights(self, filepath, by_name, skip_mismatch, options)
2266 'load_weights requires h5py when loading weights from HDF5.')
2267 if not self._is_graph_network and not self.built:
-> 2268 raise ValueError(
2269 'Unable to load weights saved in HDF5 format into a subclassed '
2270 'Model which has not created its variables yet. Call the Model '

ValueError: Unable to load weights saved in HDF5 format into a subclassed Model which has not created its variables yet. Call the Model first, then load the weights.

Any ideas for a fix?

Best regards,
Patrick

[cia-rana]

• branch: tensorflow_2
• platform: Google Colab

I encountered the same problem, and I solved it by setting self.built = True after initializing VAEModel.

class VAEModel(Model):
    def __init__(self, encoder, decoder, r_loss_factor, **kwargs):
        super(VAEModel, self).__init__(**kwargs)
        self.encoder = encoder
        self.decoder = decoder
        self.r_loss_factor = r_loss_factor
        self.built = True # added