Post-training integer quantization with int16 activations

Stay organized with collections Save and categorize content based on your preferences.

View on Run in Google Colab View source on GitHub Download notebook


TensorFlow Lite now supports converting activations to 16-bit integer values and weights to 8-bit integer values during model conversion from TensorFlow to TensorFlow Lite's flat buffer format. We refer to this mode as the "16x8 quantization mode". This mode can improve accuracy of the quantized model significantly, when activations are sensitive to the quantization, while still achieving almost 3-4x reduction in model size. Moreover, this fully quantized model can be consumed by integer-only hardware accelerators.

Some examples of models that benefit from this mode of the post-training quantization include:

  • super-resolution,
  • audio signal processing such as noise cancelling and beamforming,
  • image de-noising,
  • HDR reconstruction from a single image

In this tutorial, you train an MNIST model from scratch, check its accuracy in TensorFlow, and then convert the model into a Tensorflow Lite flatbuffer using this mode. At the end you check the accuracy of the converted model and compare it to the original float32 model. Note that this example demonstrates the usage of this mode and doesn't show benefits over other available quantization techniques in TensorFlow Lite.

Build an MNIST model


import logging

import tensorflow as tf
from tensorflow import keras
import numpy as np
import pathlib
2022-08-31 11:21:43.797906: E tensorflow/stream_executor/cuda/] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
2022-08-31 11:21:44.570625: W tensorflow/stream_executor/platform/default/] Could not load dynamic library ''; dlerror: cannot open shared object file: No such file or directory
2022-08-31 11:21:44.570882: W tensorflow/stream_executor/platform/default/] Could not load dynamic library ''; dlerror: cannot open shared object file: No such file or directory
2022-08-31 11:21:44.570895: W tensorflow/compiler/tf2tensorrt/utils/] TF-TRT Warning: Cannot dlopen some TensorRT libraries. If you would like to use Nvidia GPU with TensorRT, please make sure the missing libraries mentioned above are installed properly.

Check that the 16x8 quantization mode is available


Train and export the model

# Load MNIST dataset
mnist = keras.datasets.mnist
(train_images, train_labels), (test_images, test_labels) = mnist.load_data()

# Normalize the input image so that each pixel value is between 0 to 1.
train_images = train_images / 255.0
test_images = test_images / 255.0

# Define the model architecture
model = keras.Sequential([
  keras.layers.InputLayer(input_shape=(28, 28)),
  keras.layers.Reshape(target_shape=(28, 28, 1)),
  keras.layers.Conv2D(filters=12, kernel_size=(3, 3), activation=tf.nn.relu),
  keras.layers.MaxPooling2D(pool_size=(2, 2)),

# Train the digit classification model
  validation_data=(test_images, test_labels)
1875/1875 [==============================] - 10s 3ms/step - loss: 0.2771 - accuracy: 0.9224 - val_loss: 0.1221 - val_accuracy: 0.9660
<keras.callbacks.History at 0x7f8ede7ed760>

For the example, you trained the model for just a single epoch, so it only trains to ~96% accuracy.

Convert to a TensorFlow Lite model

Using the TensorFlow Lite Converter, you can now convert the trained model into a TensorFlow Lite model.

Now, convert the model using TFliteConverter into default float32 format:

converter = tf.lite.TFLiteConverter.from_keras_model(model)
tflite_model = converter.convert()
WARNING:absl:Found untraced functions such as _jit_compiled_convolution_op while saving (showing 1 of 1). These functions will not be directly callable after loading.
INFO:tensorflow:Assets written to: /tmpfs/tmp/tmp44cm3bqi/assets
INFO:tensorflow:Assets written to: /tmpfs/tmp/tmp44cm3bqi/assets
2022-08-31 11:22:00.729026: W tensorflow/compiler/mlir/lite/python/] Ignored output_format.
2022-08-31 11:22:00.729085: W tensorflow/compiler/mlir/lite/python/] Ignored drop_control_dependency.

Write it out to a .tflite file:

tflite_models_dir = pathlib.Path("/tmp/mnist_tflite_models/")
tflite_models_dir.mkdir(exist_ok=True, parents=True)
tflite_model_file = tflite_models_dir/"mnist_model.tflite"

To instead quantize the model to 16x8 quantization mode, first set the optimizations flag to use default optimizations. Then specify that 16x8 quantization mode is the required supported operation in the target specification:

converter.optimizations = [tf.lite.Optimize.DEFAULT]
converter.target_spec.supported_ops = [tf.lite.OpsSet.EXPERIMENTAL_TFLITE_BUILTINS_ACTIVATIONS_INT16_WEIGHTS_INT8]

As in the case of int8 post-training quantization, it is possible to produce a fully integer quantized model by setting converter options inference_input(output)_type to tf.int16.

Set the calibration data:

mnist_train, _ = tf.keras.datasets.mnist.load_data()
images = tf.cast(mnist_train[0], tf.float32) / 255.0
mnist_ds =
def representative_data_gen():
  for input_value in mnist_ds.take(100):
    # Model has only one input so each data point has one element.
    yield [input_value]
converter.representative_dataset = representative_data_gen

Finally, convert the model as usual. Note, by default the converted model will still use float input and outputs for invocation convenience.

tflite_16x8_model = converter.convert()
tflite_model_16x8_file = tflite_models_dir/"mnist_model_quant_16x8.tflite"
WARNING:absl:Found untraced functions such as _jit_compiled_convolution_op while saving (showing 1 of 1). These functions will not be directly callable after loading.
INFO:tensorflow:Assets written to: /tmpfs/tmp/tmp6huv6qqa/assets
INFO:tensorflow:Assets written to: /tmpfs/tmp/tmp6huv6qqa/assets
2022-08-31 11:22:02.184643: W tensorflow/compiler/mlir/lite/python/] Ignored output_format.
2022-08-31 11:22:02.184686: W tensorflow/compiler/mlir/lite/python/] Ignored drop_control_dependency.

Note how the resulting file is approximately 1/3 the size.

ls -lh {tflite_models_dir}
total 112K
-rw-rw-r-- 1 kbuilder kbuilder 83K Aug 31 11:22 mnist_model.tflite
-rw-rw-r-- 1 kbuilder kbuilder 25K Aug 31 11:22 mnist_model_quant_16x8.tflite

Run the TensorFlow Lite models

Run the TensorFlow Lite model using the Python TensorFlow Lite Interpreter.

Load the model into the interpreters

interpreter = tf.lite.Interpreter(model_path=str(tflite_model_file))
INFO: Created TensorFlow Lite XNNPACK delegate for CPU.
interpreter_16x8 = tf.lite.Interpreter(model_path=str(tflite_model_16x8_file))

Test the models on one image

test_image = np.expand_dims(test_images[0], axis=0).astype(np.float32)

input_index = interpreter.get_input_details()[0]["index"]
output_index = interpreter.get_output_details()[0]["index"]

interpreter.set_tensor(input_index, test_image)
predictions = interpreter.get_tensor(output_index)
import matplotlib.pylab as plt

template = "True:{true}, predicted:{predict}"
_ = plt.title(template.format(true= str(test_labels[0]),


test_image = np.expand_dims(test_images[0], axis=0).astype(np.float32)

input_index = interpreter_16x8.get_input_details()[0]["index"]
output_index = interpreter_16x8.get_output_details()[0]["index"]

interpreter_16x8.set_tensor(input_index, test_image)
predictions = interpreter_16x8.get_tensor(output_index)
template = "True:{true}, predicted:{predict}"
_ = plt.title(template.format(true= str(test_labels[0]),


Evaluate the models

# A helper function to evaluate the TF Lite model using "test" dataset.
def evaluate_model(interpreter):
  input_index = interpreter.get_input_details()[0]["index"]
  output_index = interpreter.get_output_details()[0]["index"]

  # Run predictions on every image in the "test" dataset.
  prediction_digits = []
  for test_image in test_images:
    # Pre-processing: add batch dimension and convert to float32 to match with
    # the model's input data format.
    test_image = np.expand_dims(test_image, axis=0).astype(np.float32)
    interpreter.set_tensor(input_index, test_image)

    # Run inference.

    # Post-processing: remove batch dimension and find the digit with highest
    # probability.
    output = interpreter.tensor(output_index)
    digit = np.argmax(output()[0])

  # Compare prediction results with ground truth labels to calculate accuracy.
  accurate_count = 0
  for index in range(len(prediction_digits)):
    if prediction_digits[index] == test_labels[index]:
      accurate_count += 1
  accuracy = accurate_count * 1.0 / len(prediction_digits)

  return accuracy

Repeat the evaluation on the 16x8 quantized model:

# NOTE: This quantization mode is an experimental post-training mode,
# it does not have any optimized kernels implementations or
# specialized machine learning hardware accelerators. Therefore,
# it could be slower than the float interpreter.

In this example, you have quantized a model to 16x8 with no difference in the accuracy, but with the 3x reduced size.