2

I'm want to build a sequence to sequence auto encoder for signal compression. I wanted to start with a std, LSTM based auto encoder. However, Keras complains about my model. any hint what I'm doing wrong

from keras.layers import Input, LSTM, RepeatVector
from keras.models import Model


timesteps = 10
input_dim = 4

latent_dim = 128

#Create the encoder:
inputs = Input(shape=(timesteps, input_dim))
encoded = LSTM(latent_dim)(inputs)
encoder = Model(inputs, encoded)

#Create the decoder:
decInput = Input(shape=(latent_dim))    
decoded = RepeatVector(timesteps)(decInput)
decoded = LSTM(input_dim, return_sequences=True)(decoded)
decoder = Model(decInput,decoded)

#Joining models:
joinedInput = Input(shape=(timesteps, input_dim))
encoderOut = encoder(joinedInput)    
joinedOut = decoder(encoderOut)
sequence_autoencoder = Model(joinedInput,joinedOut)

I get on the line encoded = LSTM(latent_dim)(inputs)

The error

TypeError: Expected int32, got list containing Tensors of type '_Message' instead.

CAFEBABE
  • 3,983
  • 1
  • 19
  • 38

0 Answers0