Conditional VAE - unable to build model-CodePudding

I'm currently having a VAE working with data_x dim = [300000,60,5]. I want to add data_y dim = [300000,1], which is a binary conditioner. However, I keep getting the following error:

ValueError: Graph disconnected: cannot obtain value for tensor Tensor("input_2:0", shape=(None, 1), dtype=float32) at layer "concatenate_1". The following previous layers were accessed without issue: []

I'm pretty sure concatenating the latent inputs with input_y goes wrong somehow, but I can't find the actual solution. I will post my model code below (I use a custom train_step). Maybe my entire approach is wrong?

import numpy as np
import tensorflow as tf
from tensorflow.keras.callbacks import TensorBoard

"""
## Create a sampling layer
"""

class Sampling(tf.keras.layers.Layer):
    """Uses (z_mean, z_log_var) to sample z."""

    def call(self, inputs):
        z_mean, z_log_var = inputs
        batch = tf.shape(z_mean)[0]
        dim = tf.shape(z_mean)[1]
        epsilon = tf.keras.backend.random_normal(shape=(batch, dim))
        return z_mean   tf.exp(0.5 * z_log_var) * epsilon

"""
## Build the encoder
"""

latent_dim = 8
N_chan= 5

input_x = tf.keras.layers.Input(shape=(60,N_chan))
input_y = tf.keras.layers.Input(shape=(1,))

x = tf.keras.layers.Masking(mask_value=999)(input_x)
x = tf.keras.layers.Conv1D(32, 3, activation="relu", strides=2, padding="same")(x)
x = tf.keras.layers.Conv1D(64, 3, activation="relu", strides=2, padding="same")(x)
x = tf.keras.layers.Conv1D(64, 3, activation="relu", strides=1, padding="same")(x)
x = tf.keras.layers.Conv1D(64, 3, activation="relu", strides=1, padding="same")(x)
x = tf.keras.layers.Flatten()(x)
conc = tf.keras.layers.Concatenate()([x, input_y])
x = tf.keras.layers.Dense(64, activation='relu')(conc)
# x = tf.keras.layers.Dense(latent_dim)(x)

z_mean = tf.keras.layers.Dense(latent_dim, name="z_mean")(x)
z_log_var = tf.keras.layers.Dense(latent_dim, name="z_log_var")(x)
z = Sampling()([z_mean, z_log_var])

encoder = tf.keras.Model(inputs=[input_x,input_y], outputs=[z_mean, z_log_var, z], name="encoder")
encoder.summary()

embedd = tf.keras.layers.Input(shape=(latent_dim,))
merged_input = tf.keras.layers.Concatenate()([embedd, input_y])

x = tf.keras.layers.Dense(15 * 64, activation='relu')(merged_input)
x = tf.keras.layers.Reshape(target_shape=(15, 64))(x)
x = tf.keras.layers.Conv1DTranspose(filters=64,kernel_size=3,strides=1,padding='same', activation='relu')(x)
x = tf.keras.layers.Conv1DTranspose(filters=64,kernel_size=3,strides=2,padding='same',activation='relu')(x)
x = tf.keras.layers.Conv1DTranspose(filters=64,kernel_size=3,strides=1,padding='same',activation='relu')(x)
x = tf.keras.layers.Conv1DTranspose(filters=32,kernel_size=3,strides=2,padding='same',activation="relu")(x)
decoder_outputs = tf.keras.layers.Conv1DTranspose(filters=N_chan,kernel_size=3,activation="sigmoid",padding="same")(x)

decoder = tf.keras.Model(inputs=embedd, outputs=decoder_outputs, name="decoder")
decoder.summary()

"""
## Define the VAE as a `Model` with a custom `train_step`
"""

class VAE(tf.keras.Model):
    def __init__(self, encoder, decoder, **kwargs):
        super(VAE, self).__init__(**kwargs)
        self.encoder = encoder
        self.decoder = decoder
        self.total_loss_tracker = tf.keras.metrics.Mean(name="total_loss")
        self.reconstruction_loss_tracker = tf.keras.metrics.Mean(
            name="reconstruction_loss"
        )
        self.kl_loss_tracker = tf.keras.metrics.Mean(name="kl_loss")
        self.kl_weight = 0.002

    @property
    def metrics(self):
        return [
            self.total_loss_tracker,
            self.reconstruction_loss_tracker,
            self.kl_loss_tracker,
        ]

    def train_step(self, data):
        with tf.GradientTape() as tape:
            z_mean, z_log_var, z = self.encoder(data)
            reconstruction = self.decoder(z)
            reconstruction_loss = tf.reduce_mean(
                tf.reduce_sum(
                    tf.keras.losses.binary_crossentropy(data, reconstruction), axis=(1)
                )
            )
            kl_loss = -0.5 * (1   z_log_var - tf.square(z_mean) - tf.exp(z_log_var))
            kl_loss = tf.reduce_mean(tf.reduce_sum(kl_loss, axis=1))
            total_loss = reconstruction_loss   self.kl_weight*kl_loss
        grads = tape.gradient(total_loss, self.trainable_weights)
        self.optimizer.apply_gradients(zip(grads, self.trainable_weights))
        self.total_loss_tracker.update_state(total_loss)
        self.reconstruction_loss_tracker.update_state(reconstruction_loss)
        self.kl_loss_tracker.update_state(kl_loss)
        return {
            "loss": self.total_loss_tracker.result(),
            "reconstruction_loss": self.reconstruction_loss_tracker.result(),
            "kl_loss": self.kl_loss_tracker.result(),
        }

#%% load data
file_x = "x_cvae.npy"
file_y = "y_cvae_binary.npy"
data_x,data_y = np.load(file_x),np.expand_dims(np.load(file_y),-1)

#%%
NAME = "Test"
vae = VAE(encoder, decoder)
vae.compile(optimizer=tf.keras.optimizers.Adam())
tensorboard = TensorBoard(log_dir='logs\\{}'.format(NAME),profile_batch=0)
vae.fit([data_x,data_y],data_x, epochs=10, batch_size=64
        ,validation_data=([data_x,data_y],data_x),callbacks=[tensorboard])```

CodePudding user response：

The problem is that your encoder and decoder models are sharing the same input input_y although they are two separate models. You will have to define a separate input_y for your decoder or merge the encoder and decoder into one model. Here is an example of how to use two inputs for both the encoder and the decoder.