Converting tensorflow keras LSTM model into .tflite or any working format

Question

I have trouble transforming my model into a one that can be deployed on android app via android-studio. My model is construed as follows:

model = Sequential()
model.add(LSTM(70, activation='relu', input_shape=(n_steps, n_features)))
model.add(Dense(1))
model.compile(optimizer='adam', loss='mse')

Then saved by .save function:

model.save("my.h5")

How can I open this model on android app? I have seen that tflite has problems with LSTM and my attempts to load it with dl4j failed.

Did you try loading the .pb model with TensorFlow Mobile ( not TF Lite )? It supports all the ops — Shubham Panchal, Sep 28 '19 at 00:18
No. I have seen some yt comparision but i wasnt able to find any documentation on mobile or way of adding dependency. — Sirko, Sep 28 '19 at 11:44
You can follow this [blog](https://code.tutsplus.com/tutorials/how-to-use-tensorflow-mobile-in-android-apps--cms-30957). — Shubham Panchal, Sep 28 '19 at 15:11

score 0 · Answer 1 · answered Oct 15 '19 at 10:02

0

Use the Tfite converter from tensorflow contrib(lite.TFLiteConverter.from_keras_model_file( 'model.h5')) to convert HDF5 to Tflite model.

answered Oct 15 '19 at 10:02

Siddharth Jadhav

514
3
3

Converting tensorflow keras LSTM model into .tflite or any working format

1 Answers1