Getting an error when creating the .tflite file

Hello,

I am getting an error while creating the tflite file in the following line of the code:

tflite_model = converter.convert()

The error is: tensorflow/lite/kernels/quantize.cc:110 affine_quantization->scale->size == 1 was not true.Node number 0 (QUANTIZE) failed to prepare.

When saving the Keras model, there is an input layer created, which does not have the scale and offset for quantizing, I have tried different methods but it is generated automatically. Any help is much appreciated.

TensorFlow Model Optimization version (installed from source or binary): ‘2.3.0-dev20200528’

Python version: 3

Best, Marjan

[marjanemd]

@nutsiepully I am just using the keras model for tflite converter as:

converter=tf.lite.TFLiteConverter.from_keras_model(model). Following shows the keras model, and as you can see there are two cells as InputLayer and QuantizeLayer which do not have the quantization values. I know that the issue is comming from here, but not sure how to prevent this. The model seems right to me and I don’t get the reason of first 2 cells being created.

The model is build as the following code:

model= Sequential([ quantize.quantize_annotate_layer(Conv2D(64,1,input_shape=(55,40,1),use_bias=True,activation="relu")), MaxPooling2D(2,2), Conv2D(64,(1,1),use_bias=True,activation="relu"), Flatten(), Dense(2,use_bias=True), Softmax() ])

model = quantize.quantize_apply(model)

@alanchiao As I had mentioned before my TF version is : 2.3.0

[teijeong]

Hi @marjanemd , closing this as it seems to be resolved. please reopen or file a new one if this is still a problem.