Quantize a tf.keras model that has been annotated for quantization.

Used in the notebooks

Used in the guide

Quantization constructs a model which emulates quantization during training. This allows the model to learn parameters robust to quantization loss, and also model the accuracy of a quantized model.

For more information, see

This function takes a tf.keras model in which the desired layers for quantization have already been annotated. See quantize_annotate_model and quantize_annotate_layer.


model = keras.Sequential([
    layers.Dense(10, activation='relu', input_shape=(100,)),
    quantize_annotate_layer(layers.Dense(2, activation='sigmoid'))

# Only the second Dense layer is quantized.
quantized_model = quantize_apply(model)

Note that this function removes the optimizer from the original model.

The returned model copies over weights from the original model. So while it preserves the original weights, training it will not modify the weights of the original model.

model A tf.keras Sequential or Functional model which has been annotated with quantize_annotate. It can have pre-trained weights.
scheme A QuantizeScheme which specifies transformer and quantization registry. The default is Default8BitQuantizeScheme().
quantized_layer_name_prefix A name prefix for quantized layers. The default is quant_.

Returns a new tf.keras model in which the annotated layers have been prepared for quantization.