I want to optimized the my Tensor flow model (mars-small128.pb) I have saved_model_dir directory which contain mars-small128.pb file Here is my code import tensorflow as tf converter = tf.contrib.lite.TocoConverter.from_saved_model(saved_model_dir) converter.post_training_quantize = True tflite_quantized_model = converter.convert() open("quantized_model.tflite", "wb").write(tflite_quantized_model) Error OSError: SavedModel file does not exist at: saved_model_dir/{saved_model.pbtxt|saved_model.pb} System information Os:linux 16 Python:3.5 tensorflow Installation:pip install -U tf-nightly (latest versiopn) using Cpu
Thanks
@deangelacgn This looks like my problem or very close. I just don't get it completely. I use uber's framework called ludwig. It use tf under the hood. They say that models generated by ludwig saved by tf's saver.
Directory with model looks like that:
So in documentation of this saver written:
Looks like i have variables and don't have .pb file.
So is there a way to convert model as i have it to tflite model? I'm not sure where is a gap. Is it ludwig, tflite converter or just my understanding?