View on TensorFlow.org | Run in Google Colab | View source on GitHub | Download notebook |
Overview
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
Setup
import logging
logging.getLogger("tensorflow").setLevel(logging.DEBUG)
import tensorflow as tf
from tensorflow import keras
import numpy as np
import pathlib
2024-07-19 11:32:57.822843: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:485] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered 2024-07-19 11:32:57.843687: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:8454] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered 2024-07-19 11:32:57.850049: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1452] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
Check that the 16x8 quantization mode is available
tf.lite.OpsSet.EXPERIMENTAL_TFLITE_BUILTINS_ACTIVATIONS_INT16_WEIGHTS_INT8
<OpsSet.EXPERIMENTAL_TFLITE_BUILTINS_ACTIVATIONS_INT16_WEIGHTS_INT8: 'EXPERIMENTAL_TFLITE_BUILTINS_ACTIVATIONS_INT16_WEIGHTS_INT8'>
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)),
keras.layers.Flatten(),
keras.layers.Dense(10)
])
# Train the digit classification model
model.compile(optimizer='adam',
loss=keras.losses.SparseCategoricalCrossentropy(from_logits=True),
metrics=['accuracy'])
model.fit(
train_images,
train_labels,
epochs=1,
validation_data=(test_images, test_labels)
)
/tmpfs/src/tf_docs_env/lib/python3.9/site-packages/keras/src/layers/core/input_layer.py:26: UserWarning: Argument `input_shape` is deprecated. Use `shape` instead. warnings.warn( WARNING: All log messages before absl::InitializeLog() is called are written to STDERR I0000 00:00:1721388780.938871 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388780.942464 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388780.946054 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388780.949871 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388780.961762 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388780.965081 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388780.968429 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388780.972028 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388780.975511 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388780.978730 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388780.982202 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388780.985726 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.214333 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.216311 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.218261 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.220410 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.222451 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.224283 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.226131 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.228124 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.230087 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.231904 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.233758 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.235733 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.273081 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.274988 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.276869 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.278885 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.280861 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.282685 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.284544 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.286510 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.288496 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.290804 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.293035 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 I0000 00:00:1721388782.295422 24036 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355 WARNING: All log messages before absl::InitializeLog() is called are written to STDERR I0000 00:00:1721388783.847182 24201 service.cc:146] XLA service 0x7f53f80085f0 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices: I0000 00:00:1721388783.847231 24201 service.cc:154] StreamExecutor device (0): Tesla T4, Compute Capability 7.5 I0000 00:00:1721388783.847235 24201 service.cc:154] StreamExecutor device (1): Tesla T4, Compute Capability 7.5 I0000 00:00:1721388783.847238 24201 service.cc:154] StreamExecutor device (2): Tesla T4, Compute Capability 7.5 I0000 00:00:1721388783.847241 24201 service.cc:154] StreamExecutor device (3): Tesla T4, Compute Capability 7.5 110/1875 ━━━━━━━━━━━━━━━━━━━━ 2s 1ms/step - accuracy: 0.5618 - loss: 1.6029 I0000 00:00:1721388785.025920 24201 device_compiler.h:188] Compiled cluster using XLA! This line is logged at most once for the lifetime of the process. 1875/1875 ━━━━━━━━━━━━━━━━━━━━ 6s 2ms/step - accuracy: 0.8590 - loss: 0.5079 - val_accuracy: 0.9619 - val_loss: 0.1357 <keras.src.callbacks.history.History at 0x7f55be0c2f70>
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()
INFO:tensorflow:Assets written to: /tmpfs/tmp/tmp3cpnmhnl/assets INFO:tensorflow:Assets written to: /tmpfs/tmp/tmp3cpnmhnl/assets Saved artifact at '/tmpfs/tmp/tmp3cpnmhnl'. The following endpoints are available: * Endpoint 'serve' args_0 (POSITIONAL_ONLY): TensorSpec(shape=(None, 28, 28), dtype=tf.float32, name='keras_tensor') Output Type: TensorSpec(shape=(None, 10), dtype=tf.float32, name=None) Captures: 140006282905200: TensorSpec(shape=(), dtype=tf.resource, name=None) 140006282905024: TensorSpec(shape=(), dtype=tf.resource, name=None) 140006283064064: TensorSpec(shape=(), dtype=tf.resource, name=None) 140006283063888: TensorSpec(shape=(), dtype=tf.resource, name=None) W0000 00:00:1721388789.159487 24036 tf_tfl_flatbuffer_helpers.cc:392] Ignored output_format. W0000 00:00:1721388789.159516 24036 tf_tfl_flatbuffer_helpers.cc:395] 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"
tflite_model_file.write_bytes(tflite_model)
84668
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 = tf.data.Dataset.from_tensor_slices((images)).batch(1)
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"
tflite_model_16x8_file.write_bytes(tflite_16x8_model)
INFO:tensorflow:Assets written to: /tmpfs/tmp/tmp9qbz4sxj/assets INFO:tensorflow:Assets written to: /tmpfs/tmp/tmp9qbz4sxj/assets Saved artifact at '/tmpfs/tmp/tmp9qbz4sxj'. The following endpoints are available: * Endpoint 'serve' args_0 (POSITIONAL_ONLY): TensorSpec(shape=(None, 28, 28), dtype=tf.float32, name='keras_tensor') Output Type: TensorSpec(shape=(None, 10), dtype=tf.float32, name=None) Captures: 140006282905200: TensorSpec(shape=(), dtype=tf.resource, name=None) 140006282905024: TensorSpec(shape=(), dtype=tf.resource, name=None) 140006283064064: TensorSpec(shape=(), dtype=tf.resource, name=None) 140006283063888: TensorSpec(shape=(), dtype=tf.resource, name=None) W0000 00:00:1721388790.014921 24036 tf_tfl_flatbuffer_helpers.cc:392] Ignored output_format. W0000 00:00:1721388790.014945 24036 tf_tfl_flatbuffer_helpers.cc:395] Ignored drop_control_dependency. 25144
Note how the resulting file is approximately 1/3
the size.
ls -lh {tflite_models_dir}
total 140K -rw-rw-r-- 1 kbuilder kbuilder 83K Jul 19 11:33 mnist_model.tflite -rw-rw-r-- 1 kbuilder kbuilder 25K Jul 19 11:31 mnist_model_quant.tflite -rw-rw-r-- 1 kbuilder kbuilder 25K Jul 19 11:33 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))
interpreter.allocate_tensors()
INFO: Created TensorFlow Lite XNNPACK delegate for CPU.
interpreter_16x8 = tf.lite.Interpreter(model_path=str(tflite_model_16x8_file))
interpreter_16x8.allocate_tensors()
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)
interpreter.invoke()
predictions = interpreter.get_tensor(output_index)
import matplotlib.pylab as plt
plt.imshow(test_images[0])
template = "True:{true}, predicted:{predict}"
_ = plt.title(template.format(true= str(test_labels[0]),
predict=str(np.argmax(predictions[0]))))
plt.grid(False)
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)
interpreter_16x8.invoke()
predictions = interpreter_16x8.get_tensor(output_index)
plt.imshow(test_images[0])
template = "True:{true}, predicted:{predict}"
_ = plt.title(template.format(true= str(test_labels[0]),
predict=str(np.argmax(predictions[0]))))
plt.grid(False)
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.
interpreter.invoke()
# Post-processing: remove batch dimension and find the digit with highest
# probability.
output = interpreter.tensor(output_index)
digit = np.argmax(output()[0])
prediction_digits.append(digit)
# 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
print(evaluate_model(interpreter))
0.9619
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.
print(evaluate_model(interpreter_16x8))
0.9618
In this example, you have quantized a model to 16x8 with no difference in the accuracy, but with the 3x reduced size.