Kuantisasi rentang dinamis pasca-pelatihan

Lihat di TensorFlow.org Jalankan di Google Colab Lihat sumber di GitHub Unduh buku catatan Lihat model TF Hub

Ringkasan

TensorFlow Lite sekarang mendukung mengkonversi bobot untuk 8 bit presisi sebagai bagian dari konversi Model dari graphdefs tensorflow ke format penyangga datar TensorFlow Lite. Kuantisasi rentang dinamis mencapai pengurangan 4x dalam ukuran model. Selain itu, TFLite mendukung kuantisasi dan dekuantisasi aktivasi dengan cepat untuk memungkinkan:

  1. Menggunakan kernel terkuantisasi untuk implementasi yang lebih cepat jika tersedia.
  2. Pencampuran kernel floating-point dengan kernel terkuantisasi untuk berbagai bagian grafik.

Aktivasi selalu disimpan dalam floating point. Untuk operasi yang mendukung kernel terkuantisasi, aktivasi dikuantisasi hingga 8 bit presisi secara dinamis sebelum pemrosesan dan didekuantisasi untuk mengapung presisi setelah pemrosesan. Bergantung pada model yang dikonversi, ini dapat memberikan percepatan pada komputasi floating point murni.

Berbeda dengan kuantisasi pelatihan sadar , bobot yang pelatihan pasca terkuantisasi dan aktivasi dikuantisasi dinamis di inferensi dalam metode ini. Oleh karena itu, bobot model tidak dilatih ulang untuk mengkompensasi kesalahan yang diinduksi kuantisasi. Penting untuk memeriksa keakuratan model terkuantisasi untuk memastikan bahwa degradasi dapat diterima.

Tutorial ini melatih model MNIST dari awal, memeriksa keakuratannya di TensorFlow, lalu mengonversi model menjadi flatbuffer Tensorflow Lite dengan kuantisasi rentang dinamis. Akhirnya, ia memeriksa keakuratan model yang dikonversi dan membandingkannya dengan model float asli.

Bangun model MNIST

Mempersiapkan

import logging
logging.getLogger("tensorflow").setLevel(logging.DEBUG)

import tensorflow as tf
from tensorflow import keras
import numpy as np
import pathlib

Latih model TensorFlow

# 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)
)
1875/1875 [==============================] - 6s 2ms/step - loss: 0.3260 - accuracy: 0.9063 - val_loss: 0.1721 - val_accuracy: 0.9499
<keras.callbacks.History at 0x7fb7a1c4ed90>

Misalnya, karena Anda melatih model hanya untuk satu epoch, jadi model ini hanya melatih akurasi ~96%.

Konversikan ke model TensorFlow Lite

Menggunakan Python TFLiteConverter , Anda sekarang dapat mengkonversi model dilatih menjadi model TensorFlow Lite.

Sekarang memuat model menggunakan TFLiteConverter :

converter = tf.lite.TFLiteConverter.from_keras_model(model)
tflite_model = converter.convert()
2021-11-02 11:23:32.211024: W tensorflow/python/util/util.cc:348] Sets are not currently considered sequences, but this may change in the future, so consider avoiding using them.
INFO:tensorflow:Assets written to: /tmp/tmpua453ven/assets
2021-11-02 11:23:32.640259: W tensorflow/compiler/mlir/lite/python/tf_tfl_flatbuffer_helpers.cc:351] Ignored output_format.
2021-11-02 11:23:32.640302: W tensorflow/compiler/mlir/lite/python/tf_tfl_flatbuffer_helpers.cc:354] Ignored drop_control_dependency.

Tulis ke file tflite:

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)
84500

Untuk quantize model pada ekspor, mengatur optimizations bendera untuk mengoptimalkan untuk ukuran:

converter.optimizations = [tf.lite.Optimize.DEFAULT]
tflite_quant_model = converter.convert()
tflite_model_quant_file = tflite_models_dir/"mnist_model_quant.tflite"
tflite_model_quant_file.write_bytes(tflite_quant_model)
INFO:tensorflow:Assets written to: /tmp/tmpaw0wsb_y/assets
INFO:tensorflow:Assets written to: /tmp/tmpaw0wsb_y/assets
2021-11-02 11:23:33.235475: W tensorflow/compiler/mlir/lite/python/tf_tfl_flatbuffer_helpers.cc:351] Ignored output_format.
2021-11-02 11:23:33.235512: W tensorflow/compiler/mlir/lite/python/tf_tfl_flatbuffer_helpers.cc:354] Ignored drop_control_dependency.
23904

Perhatikan bagaimana file yang dihasilkan, kira-kira 1/4 ukuran.

ls -lh {tflite_models_dir}
total 136K
-rw-rw-r-- 1 kbuilder kbuilder 83K Nov  2 11:23 mnist_model.tflite
-rw-rw-r-- 1 kbuilder kbuilder 24K Nov  2 11:23 mnist_model_quant.tflite
-rw-rw-r-- 1 kbuilder kbuilder 25K Nov  2 11:20 mnist_model_quant_16x8.tflite

Jalankan model TFLite

Jalankan model TensorFlow Lite menggunakan Interpreter Python TensorFlow Lite.

Muat model ke penerjemah

interpreter = tf.lite.Interpreter(model_path=str(tflite_model_file))
interpreter.allocate_tensors()
interpreter_quant = tf.lite.Interpreter(model_path=str(tflite_model_quant_file))
interpreter_quant.allocate_tensors()

Uji model pada satu gambar

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)

png

Evaluasi modelnya

# 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))
0.9499

Ulangi evaluasi pada model terkuantisasi rentang dinamis untuk mendapatkan:

print(evaluate_model(interpreter_quant))
0.95

Dalam contoh ini, model terkompresi tidak memiliki perbedaan dalam akurasi.

Mengoptimalkan model yang ada

Resnet dengan lapisan pra-aktivasi (Resnet-v2) banyak digunakan untuk aplikasi visi. Pre-dilatih grafik beku untuk resnet-v2-101 tersedia di Tensorflow Hub .

Anda dapat mengonversi grafik beku menjadi flatbuffer TensorFLow Lite dengan kuantisasi dengan:

import tensorflow_hub as hub

resnet_v2_101 = tf.keras.Sequential([
  keras.layers.InputLayer(input_shape=(224, 224, 3)),
  hub.KerasLayer("https://tfhub.dev/google/imagenet/resnet_v2_101/classification/4")
])

converter = tf.lite.TFLiteConverter.from_keras_model(resnet_v2_101)
# Convert to TF Lite without quantization
resnet_tflite_file = tflite_models_dir/"resnet_v2_101.tflite"
resnet_tflite_file.write_bytes(converter.convert())
WARNING:tensorflow:Compiled the loaded model, but the compiled metrics have yet to be built. `model.compile_metrics` will be empty until you train or evaluate the model.
WARNING:tensorflow:Compiled the loaded model, but the compiled metrics have yet to be built. `model.compile_metrics` will be empty until you train or evaluate the model.
INFO:tensorflow:Assets written to: /tmp/tmpxtji1amp/assets
INFO:tensorflow:Assets written to: /tmp/tmpxtji1amp/assets
2021-11-02 11:23:57.616139: W tensorflow/compiler/mlir/lite/python/tf_tfl_flatbuffer_helpers.cc:351] Ignored output_format.
2021-11-02 11:23:57.616201: W tensorflow/compiler/mlir/lite/python/tf_tfl_flatbuffer_helpers.cc:354] Ignored drop_control_dependency.
178509352
# Convert to TF Lite with quantization
converter.optimizations = [tf.lite.Optimize.DEFAULT]
resnet_quantized_tflite_file = tflite_models_dir/"resnet_v2_101_quantized.tflite"
resnet_quantized_tflite_file.write_bytes(converter.convert())
WARNING:tensorflow:Compiled the loaded model, but the compiled metrics have yet to be built. `model.compile_metrics` will be empty until you train or evaluate the model.
WARNING:tensorflow:Compiled the loaded model, but the compiled metrics have yet to be built. `model.compile_metrics` will be empty until you train or evaluate the model.
INFO:tensorflow:Assets written to: /tmp/tmpg169iato/assets
INFO:tensorflow:Assets written to: /tmp/tmpg169iato/assets
2021-11-02 11:24:12.965799: W tensorflow/compiler/mlir/lite/python/tf_tfl_flatbuffer_helpers.cc:351] Ignored output_format.
2021-11-02 11:24:12.965851: W tensorflow/compiler/mlir/lite/python/tf_tfl_flatbuffer_helpers.cc:354] Ignored drop_control_dependency.
46256864
ls -lh {tflite_models_dir}/*.tflite
-rw-rw-r-- 1 kbuilder kbuilder  83K Nov  2 11:23 /tmp/mnist_tflite_models/mnist_model.tflite
-rw-rw-r-- 1 kbuilder kbuilder  24K Nov  2 11:23 /tmp/mnist_tflite_models/mnist_model_quant.tflite
-rw-rw-r-- 1 kbuilder kbuilder  25K Nov  2 11:20 /tmp/mnist_tflite_models/mnist_model_quant_16x8.tflite
-rw-rw-r-- 1 kbuilder kbuilder 171M Nov  2 11:23 /tmp/mnist_tflite_models/resnet_v2_101.tflite
-rw-rw-r-- 1 kbuilder kbuilder  45M Nov  2 11:24 /tmp/mnist_tflite_models/resnet_v2_101_quantized.tflite

Ukuran model berkurang dari 171 MB menjadi 43 MB. Akurasi model ini pada imagenet dapat dievaluasi dengan menggunakan script yang disediakan untuk pengukuran akurasi TFLite .

Akurasi top-1 model yang dioptimalkan adalah 76,8, sama dengan model floating point.