Google I/O returns May 18-20! Reserve space and build your schedule Register now

Model Averaging

View on TensorFlow.org Run in Google Colab View source on GitHub Download notebook

Overview

This notebook demonstrates how to use Moving Average Optimizer along with the Model Average Checkpoint from tensorflow addons pagkage.

Moving Averaging

The advantage of Moving Averaging is that they are less prone to rampant loss shifts or irregular data representation in the latest batch. It gives a smooothened and a more genral idea of the model training until some point.

Stocastic Averaging

Stocastic Weight Averaging converges to wider optimas. By doing so, it resembles geometric ensembeling. SWA is a simple method to improve model performance when used as a wrapper around other optimizers and averaging results from different points of trajectory of the inner optimizer.

Model Average Checkpoint

callbacks.ModelCheckpoint doesn't give you the option to save moving average weights in the middle of training, which is why Model Average Optimizers required a custom callback. Using the update_weights parameter, ModelAverageCheckpoint allows you to:

  1. Assign the moving average weights to the model, and save them.
  2. Keep the old non-averaged weights, but the saved model uses the average weights.

Setup

pip install -q -U tensorflow-addons

import tensorflow as tf
import tensorflow_addons as tfa

import numpy as np
import os

Build Model

def create_model(opt):
    model = tf.keras.models.Sequential([
        tf.keras.layers.Flatten(),                         
        tf.keras.layers.Dense(64, activation='relu'),
        tf.keras.layers.Dense(64, activation='relu'),
        tf.keras.layers.Dense(10, activation='softmax')
    ])

    model.compile(optimizer=opt,
                    loss='sparse_categorical_crossentropy',
                    metrics=['accuracy'])

    return model

Prepare Dataset

#Load Fashion MNIST dataset
train, test = tf.keras.datasets.fashion_mnist.load_data()

images, labels = train
images = images/255.0
labels = labels.astype(np.int32)

fmnist_train_ds = tf.data.Dataset.from_tensor_slices((images, labels))
fmnist_train_ds = fmnist_train_ds.shuffle(5000).batch(32)

test_images, test_labels = test

Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-labels-idx1-ubyte.gz
32768/29515 [=================================] - 0s 0us/step
Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-images-idx3-ubyte.gz
26427392/26421880 [==============================] - 1s 0us/step
Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-labels-idx1-ubyte.gz
8192/5148 [===============================================] - 0s 0us/step
Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-images-idx3-ubyte.gz
4423680/4422102 [==============================] - 0s 0us/step

We will be comparing three optimizers here:

  • Unwrapped SGD
  • SGD with Moving Average
  • SGD with Stochastic Weight Averaging

And see how they perform with the same model.

#Optimizers 
sgd = tf.keras.optimizers.SGD(0.01)
moving_avg_sgd = tfa.optimizers.MovingAverage(sgd)
stocastic_avg_sgd = tfa.optimizers.SWA(sgd)

Both MovingAverage and StocasticAverage optimers use ModelAverageCheckpoint.

#Callback 
checkpoint_path = "./training/cp-{epoch:04d}.ckpt"
checkpoint_dir = os.path.dirname(checkpoint_path)

cp_callback = tf.keras.callbacks.ModelCheckpoint(filepath=checkpoint_dir,
                                                 save_weights_only=True,
                                                 verbose=1)
avg_callback = tfa.callbacks.AverageModelCheckpoint(filepath=checkpoint_dir, 
                                                    update_weights=True)

Train Model

Vanilla SGD Optimizer

#Build Model
model = create_model(sgd)

#Train the network
model.fit(fmnist_train_ds, epochs=5, callbacks=[cp_callback])

Epoch 1/5
1875/1875 [==============================] - 5s 2ms/step - loss: 1.1010 - accuracy: 0.6480

Epoch 00001: saving model to ./training
Epoch 2/5
1875/1875 [==============================] - 4s 2ms/step - loss: 0.5210 - accuracy: 0.8180

Epoch 00002: saving model to ./training
Epoch 3/5
1875/1875 [==============================] - 4s 2ms/step - loss: 0.4636 - accuracy: 0.8355

Epoch 00003: saving model to ./training
Epoch 4/5
1875/1875 [==============================] - 4s 2ms/step - loss: 0.4320 - accuracy: 0.8471

Epoch 00004: saving model to ./training
Epoch 5/5
1875/1875 [==============================] - 4s 2ms/step - loss: 0.4098 - accuracy: 0.8558

Epoch 00005: saving model to ./training
<tensorflow.python.keras.callbacks.History at 0x7f9c00068b00>

#Evalute results
model.load_weights(checkpoint_dir)
loss, accuracy = model.evaluate(test_images, test_labels, batch_size=32, verbose=2)
print("Loss :", loss)
print("Accuracy :", accuracy)

313/313 - 1s - loss: 82.8866 - accuracy: 0.7894
Loss : 82.88660430908203
Accuracy : 0.7893999814987183

Moving Average SGD

#Build Model
model = create_model(moving_avg_sgd)

#Train the network
model.fit(fmnist_train_ds, epochs=5, callbacks=[avg_callback])

Epoch 1/5
1875/1875 [==============================] - 5s 2ms/step - loss: 1.1354 - accuracy: 0.6366
INFO:tensorflow:Assets written to: ./training/assets
Epoch 2/5
1875/1875 [==============================] - 5s 2ms/step - loss: 0.5312 - accuracy: 0.8170
INFO:tensorflow:Assets written to: ./training/assets
Epoch 3/5
1875/1875 [==============================] - 5s 2ms/step - loss: 0.4677 - accuracy: 0.8366
INFO:tensorflow:Assets written to: ./training/assets
Epoch 4/5
1875/1875 [==============================] - 5s 2ms/step - loss: 0.4407 - accuracy: 0.8447
INFO:tensorflow:Assets written to: ./training/assets
Epoch 5/5
1875/1875 [==============================] - 5s 2ms/step - loss: 0.4174 - accuracy: 0.8542
INFO:tensorflow:Assets written to: ./training/assets
<tensorflow.python.keras.callbacks.History at 0x7f9c0005a358>

#Evalute results
model.load_weights(checkpoint_dir)
loss, accuracy = model.evaluate(test_images, test_labels, batch_size=32, verbose=2)
print("Loss :", loss)
print("Accuracy :", accuracy)

313/313 - 0s - loss: 82.8866 - accuracy: 0.7894
Loss : 82.88660430908203
Accuracy : 0.7893999814987183

Stocastic Weight Average SGD

#Build Model
model = create_model(stocastic_avg_sgd)

#Train the network
model.fit(fmnist_train_ds, epochs=5, callbacks=[avg_callback])

Epoch 1/5
1875/1875 [==============================] - 6s 3ms/step - loss: 1.0615 - accuracy: 0.6521
INFO:tensorflow:Assets written to: ./training/assets
Epoch 2/5
1875/1875 [==============================] - 5s 3ms/step - loss: 0.6078 - accuracy: 0.7925
INFO:tensorflow:Assets written to: ./training/assets
Epoch 3/5
1875/1875 [==============================] - 5s 3ms/step - loss: 0.5671 - accuracy: 0.8062
INFO:tensorflow:Assets written to: ./training/assets
Epoch 4/5
1875/1875 [==============================] - 5s 3ms/step - loss: 0.5468 - accuracy: 0.8115
INFO:tensorflow:Assets written to: ./training/assets
Epoch 5/5
1875/1875 [==============================] - 5s 3ms/step - loss: 0.5288 - accuracy: 0.8181
INFO:tensorflow:Assets written to: ./training/assets
<tensorflow.python.keras.callbacks.History at 0x7f9b7c6faa58>

#Evalute results
model.load_weights(checkpoint_dir)
loss, accuracy = model.evaluate(test_images, test_labels, batch_size=32, verbose=2)
print("Loss :", loss)
print("Accuracy :", accuracy)

313/313 - 0s - loss: 82.8866 - accuracy: 0.7894
Loss : 82.88660430908203
Accuracy : 0.7893999814987183