This is my LSTM model:
lstm_out1 = 150
embed_dim = 768
model = Sequential()
model.add(Embedding(embed_tensor.shape[0], embed_dim, weights=[embed_tensor], input_length=512, trainable=False))
model.add(LSTM(lstm_out1, dropout=0.2, recurrent_dropout=0.2))
model.add(Dense(64, activation='relu'))
model.add(Dense(1, activation='sigmoid'))
adam = Adam(lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-08, decay=0.0)
model.compile(loss='binary_crossentropy',
optimizer=adam,
metrics=['accuracy'])
model.summary()
Model summary is:
Model: "sequential_8"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
embedding_12 (Embedding) (None, 512, 768) 91812096
lstm_10 (LSTM) (None, 150) 551400
dense_16 (Dense) (None, 64) 9664
dense_17 (Dense) (None, 1) 65
I called model.fit() on tis:
model.fit(token_tensor, labels, batch_size=10, epochs=1, shuffle=True)
It's producing error:
ValueError: Input 0 of layer "sequential_8" is incompatible with the layer: expected shape=(None, 512), found shape=(10, 1, 512)
My each input vector is size of 512. the 10 in found shape is the batch size. Plz help me to resolve this.
As the error indicates, your input tensor shape of (10, 1, 512) is not compatible with the declared input of the model, which has shape (None, 512). The "None" in the input shape indicates that the size for that axis is arbitrary.
To fix your input, you need to remove the second dimension (the "1"), after which it will be compatible with the model. Depending on the exact type of token_tensor, you could achieve this by
token_tensor = token_tensor[:, 0, :]
or
token_tensor = token_tensor.squeeze()
or
token_tensor = token_tensor.reshape((-1, token_tensor.shape[-1]))