My model is trained on the "imdb reviews dataset" and works fine when predicting the sentiment of movie reviews. However, when I convert my model for Tensorflow Lite, it outputs: None is only supported in the 1st dimension. Tensor 'embedding 1 input' has invalid shape '[None, None]'. When training my model, I did not specify a specific shape, therefore I am unsure of what shape to pass for my model to work with my android app. (As long as I convert the embedding_input shape to something else, the TFLite model will be created, but does not work with my android app)
Code for model:
from tensorflow import keras
from keras.preprocessing.text import Tokenizer
from keras.preprocessing.sequence import pad_sequences
from keras.layers import Dense , Input , LSTM , Embedding, Dropout , Activation, GRU, Flatten
from keras.layers import Bidirectional, GlobalMaxPool1D
from keras.models import Model, Sequential
from keras.layers import Convolution1D
from keras import initializers, regularizers, constraints, optimizers, layers
max_features = 6000
tokenizer = Tokenizer(num_words=max_features)
tokenizer.fit_on_texts(df['Processed_Reviews'])
list_tokenized_train = tokenizer.texts_to_sequences(df['Processed_Reviews'])
maxlen = 130
X_t = pad_sequences(list_tokenized_train, maxlen=maxlen)
y = df['sentiment']
embed_size = 128
model = Sequential()
model.add(Embedding(max_features, embed_size))
model.add(Bidirectional(LSTM(32, return_sequences = True)))
model.add(GlobalMaxPool1D())
model.add(Dense(20, activation="relu"))
model.add(Dropout(0.05))
model.add(Dense(1, activation="sigmoid"))
model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
batch_size = 100
epochs = 3
model.fit(X_t,y, batch_size=batch_size, epochs=epochs, validation_split=0.2)
#Conversion Code
import tensorflow as tf
inference_model = tf.keras.models.load_model('imdb-reviews-final.h5')
#inference_model.input.set_shape((6000, 128)) --> Reshaping allows model conversion to happen, but does not actually work with the app
converter = tf.lite.TFLiteConverter.from_keras_model(inference_model)
tflite_model = converter.convert()
open("model.tflite", "wb").write(tflite_model)
Current stable versions of Tensorflow don't support dynamic input shapes.
However, using the nightly build could solve your problem. I found this issue in Tensorflow github where this method is discussed. However, I'm not sure if this works on Android.