Writing a training loop from scratch

Author: fchollet

View on TensorFlow.org Run in Google Colab View source on GitHub View on keras.io

Setup

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

Introduction

Keras provides default training and evaluation loops, fit() and evaluate(). Their usage is covered in the guide Training & evaluation with the built-in methods.

If you want to customize the learning algorithm of your model while still leveraging the convenience of fit() (for instance, to train a GAN using fit()), you can subclass the Model class and implement your own train_step() method, which is called repeatedly during fit(). This is covered in the guide Customizing what happens in fit().

Now, if you want very low-level control over training & evaluation, you should write your own training & evaluation loops from scratch. This is what this guide is about.

Using the GradientTape: a first end-to-end example

Calling a model inside a GradientTape scope enables you to retrieve the gradients of the trainable weights of the layer with respect to a loss value. Using an optimizer instance, you can use these gradients to update these variables (which you can retrieve using model.trainable_weights).

Let's consider a simple MNIST model:

inputs = keras.Input(shape=(784,), name="digits")
x1 = layers.Dense(64, activation="relu")(inputs)
x2 = layers.Dense(64, activation="relu")(x1)
outputs = layers.Dense(10, name="predictions")(x2)
model = keras.Model(inputs=inputs, outputs=outputs)

Let's train it using mini-batch gradient with a custom training loop.

First, we're going to need an optimizer, a loss function, and a dataset:

# Instantiate an optimizer.
optimizer = keras.optimizers.SGD(learning_rate=1e-3)
# Instantiate a loss function.
loss_fn = keras.losses.SparseCategoricalCrossentropy(from_logits=True)

# Prepare the training dataset.
batch_size = 64
(x_train, y_train), (x_test, y_test) = keras.datasets.mnist.load_data()
x_train = np.reshape(x_train, (-1, 784))
x_test = np.reshape(x_test, (-1, 784))

# Reserve 10,000 samples for validation.
x_val = x_train[-10000:]
y_val = y_train[-10000:]
x_train = x_train[:-10000]
y_train = y_train[:-10000]

# Prepare the training dataset.
train_dataset = tf.data.Dataset.from_tensor_slices((x_train, y_train))
train_dataset = train_dataset.shuffle(buffer_size=1024).batch(batch_size)

# Prepare the validation dataset.
val_dataset = tf.data.Dataset.from_tensor_slices((x_val, y_val))
val_dataset = val_dataset.batch(batch_size)

Here's our training loop:

  • We open a for loop that iterates over epochs
  • For each epoch, we open a for loop that iterates over the dataset, in batches
  • For each batch, we open a GradientTape() scope
  • Inside this scope, we call the model (forward pass) and compute the loss
  • Outside the scope, we retrieve the gradients of the weights of the model with regard to the loss
  • Finally, we use the optimizer to update the weights of the model based on the gradients
epochs = 2
for epoch in range(epochs):
    print("\nStart of epoch %d" % (epoch,))

    # Iterate over the batches of the dataset.
    for step, (x_batch_train, y_batch_train) in enumerate(train_dataset):
        # Open a GradientTape to record the operations run
        # during the forward pass, which enables auto-differentiation.
        with tf.GradientTape() as tape:
            # Run the forward pass of the layer.
            # The operations that the layer applies
            # to its inputs are going to be recorded
            # on the GradientTape.
            logits = model(x_batch_train, training=True)  # Logits for this minibatch

            # Compute the loss value for this minibatch.
            loss_value = loss_fn(y_batch_train, logits)

        # Use the gradient tape to automatically retrieve
        # the gradients of the trainable variables with respect to the loss.
        grads = tape.gradient(loss_value, model.trainable_weights)

        # Run one step of gradient descent by updating
        # the value of the variables to minimize the loss.
        optimizer.apply_gradients(zip(grads, model.trainable_weights))

        # Log every 200 batches.
        if step % 200 == 0:
            print(
                "Training loss (for one batch) at step %d: %.4f"
                % (step, float(loss_value))
            )
            print("Seen so far: %s samples" % ((step + 1) * batch_size))

Start of epoch 0
WARNING:tensorflow:5 out of the last 5 calls to <function _BaseOptimizer._update_step_xla at 0x7f51fe36a4c0> triggered tf.function retracing. Tracing is expensive and the excessive number of tracings could be due to (1) creating @tf.function repeatedly in a loop, (2) passing tensors with different shapes, (3) passing Python objects instead of tensors. For (1), please define your @tf.function outside of the loop. For (2), @tf.function has reduce_retracing=True option that can avoid unnecessary retracing. For (3), please refer to https://www.tensorflow.org/guide/function#controlling_retracing and https://www.tensorflow.org/api_docs/python/tf/function for  more details.
WARNING:tensorflow:6 out of the last 6 calls to <function _BaseOptimizer._update_step_xla at 0x7f51fe36a4c0> triggered tf.function retracing. Tracing is expensive and the excessive number of tracings could be due to (1) creating @tf.function repeatedly in a loop, (2) passing tensors with different shapes, (3) passing Python objects instead of tensors. For (1), please define your @tf.function outside of the loop. For (2), @tf.function has reduce_retracing=True option that can avoid unnecessary retracing. For (3), please refer to https://www.tensorflow.org/guide/function#controlling_retracing and https://www.tensorflow.org/api_docs/python/tf/function for  more details.
Training loss (for one batch) at step 0: 131.3794
Seen so far: 64 samples
Training loss (for one batch) at step 200: 1.2871
Seen so far: 12864 samples
Training loss (for one batch) at step 400: 1.2652
Seen so far: 25664 samples
Training loss (for one batch) at step 600: 0.8800
Seen so far: 38464 samples

Start of epoch 1
Training loss (for one batch) at step 0: 0.8296
Seen so far: 64 samples
Training loss (for one batch) at step 200: 1.3322
Seen so far: 12864 samples
Training loss (for one batch) at step 400: 1.0486
Seen so far: 25664 samples
Training loss (for one batch) at step 600: 0.6610
Seen so far: 38464 samples

Low-level handling of metrics

Let's add metrics monitoring to this basic loop.

You can readily reuse the built-in metrics (or custom ones you wrote) in such training loops written from scratch. Here's the flow:

  • Instantiate the metric at the start of the loop
  • Call metric.update_state() after each batch
  • Call metric.result() when you need to display the current value of the metric
  • Call metric.reset_states() when you need to clear the state of the metric (typically at the end of an epoch)

Let's use this knowledge to compute SparseCategoricalAccuracy on validation data at the end of each epoch:

# Get model
inputs = keras.Input(shape=(784,), name="digits")
x = layers.Dense(64, activation="relu", name="dense_1")(inputs)
x = layers.Dense(64, activation="relu", name="dense_2")(x)
outputs = layers.Dense(10, name="predictions")(x)
model = keras.Model(inputs=inputs, outputs=outputs)

# Instantiate an optimizer to train the model.
optimizer = keras.optimizers.SGD(learning_rate=1e-3)
# Instantiate a loss function.
loss_fn = keras.losses.SparseCategoricalCrossentropy(from_logits=True)

# Prepare the metrics.
train_acc_metric = keras.metrics.SparseCategoricalAccuracy()
val_acc_metric = keras.metrics.SparseCategoricalAccuracy()

Here's our training & evaluation loop:

import time

epochs = 2
for epoch in range(epochs):
    print("\nStart of epoch %d" % (epoch,))
    start_time = time.time()

    # Iterate over the batches of the dataset.
    for step, (x_batch_train, y_batch_train) in enumerate(train_dataset):
        with tf.GradientTape() as tape:
            logits = model(x_batch_train, training=True)
            loss_value = loss_fn(y_batch_train, logits)
        grads = tape.gradient(loss_value, model.trainable_weights)
        optimizer.apply_gradients(zip(grads, model.trainable_weights))

        # Update training metric.
        train_acc_metric.update_state(y_batch_train, logits)

        # Log every 200 batches.
        if step % 200 == 0:
            print(
                "Training loss (for one batch) at step %d: %.4f"
                % (step, float(loss_value))
            )
            print("Seen so far: %d samples" % ((step + 1) * batch_size))

    # Display metrics at the end of each epoch.
    train_acc = train_acc_metric.result()
    print("Training acc over epoch: %.4f" % (float(train_acc),))

    # Reset training metrics at the end of each epoch
    train_acc_metric.reset_states()

    # Run a validation loop at the end of each epoch.
    for x_batch_val, y_batch_val in val_dataset:
        val_logits = model(x_batch_val, training=False)
        # Update val metrics
        val_acc_metric.update_state(y_batch_val, val_logits)
    val_acc = val_acc_metric.result()
    val_acc_metric.reset_states()
    print("Validation acc: %.4f" % (float(val_acc),))
    print("Time taken: %.2fs" % (time.time() - start_time))

Start of epoch 0
Training loss (for one batch) at step 0: 106.2691
Seen so far: 64 samples
Training loss (for one batch) at step 200: 0.9259
Seen so far: 12864 samples
Training loss (for one batch) at step 400: 0.9347
Seen so far: 25664 samples
Training loss (for one batch) at step 600: 0.7641
Seen so far: 38464 samples
Training acc over epoch: 0.7332
Validation acc: 0.8325
Time taken: 10.95s

Start of epoch 1
Training loss (for one batch) at step 0: 0.5238
Seen so far: 64 samples
Training loss (for one batch) at step 200: 0.7125
Seen so far: 12864 samples
Training loss (for one batch) at step 400: 0.5705
Seen so far: 25664 samples
Training loss (for one batch) at step 600: 0.6006
Seen so far: 38464 samples
Training acc over epoch: 0.8424
Validation acc: 0.8525
Time taken: 10.59s

Speeding-up your training step with tf.function

The default runtime in TensorFlow 2 is eager execution. As such, our training loop above executes eagerly.

This is great for debugging, but graph compilation has a definite performance advantage. Describing your computation as a static graph enables the framework to apply global performance optimizations. This is impossible when the framework is constrained to greedily execute one operation after another, with no knowledge of what comes next.

You can compile into a static graph any function that takes tensors as input. Just add a @tf.function decorator on it, like this:

@tf.function
def train_step(x, y):
    with tf.GradientTape() as tape:
        logits = model(x, training=True)
        loss_value = loss_fn(y, logits)
    grads = tape.gradient(loss_value, model.trainable_weights)
    optimizer.apply_gradients(zip(grads, model.trainable_weights))
    train_acc_metric.update_state(y, logits)
    return loss_value

Let's do the same with the evaluation step:

@tf.function
def test_step(x, y):
    val_logits = model(x, training=False)
    val_acc_metric.update_state(y, val_logits)

Now, let's re-run our training loop with this compiled training step:

import time

epochs = 2
for epoch in range(epochs):
    print("\nStart of epoch %d" % (epoch,))
    start_time = time.time()

    # Iterate over the batches of the dataset.
    for step, (x_batch_train, y_batch_train) in enumerate(train_dataset):
        loss_value = train_step(x_batch_train, y_batch_train)

        # Log every 200 batches.
        if step % 200 == 0:
            print(
                "Training loss (for one batch) at step %d: %.4f"
                % (step, float(loss_value))
            )
            print("Seen so far: %d samples" % ((step + 1) * batch_size))

    # Display metrics at the end of each epoch.
    train_acc = train_acc_metric.result()
    print("Training acc over epoch: %.4f" % (float(train_acc),))

    # Reset training metrics at the end of each epoch
    train_acc_metric.reset_states()

    # Run a validation loop at the end of each epoch.
    for x_batch_val, y_batch_val in val_dataset:
        test_step(x_batch_val, y_batch_val)

    val_acc = val_acc_metric.result()
    val_acc_metric.reset_states()
    print("Validation acc: %.4f" % (float(val_acc),))
    print("Time taken: %.2fs" % (time.time() - start_time))

Start of epoch 0
Training loss (for one batch) at step 0: 0.5162
Seen so far: 64 samples
Training loss (for one batch) at step 200: 0.4599
Seen so far: 12864 samples
Training loss (for one batch) at step 400: 0.3975
Seen so far: 25664 samples
Training loss (for one batch) at step 600: 0.2557
Seen so far: 38464 samples
Training acc over epoch: 0.8747
Validation acc: 0.8545
Time taken: 1.85s

Start of epoch 1
Training loss (for one batch) at step 0: 0.6145
Seen so far: 64 samples
Training loss (for one batch) at step 200: 0.3751
Seen so far: 12864 samples
Training loss (for one batch) at step 400: 0.3464
Seen so far: 25664 samples
Training loss (for one batch) at step 600: 0.4128
Seen so far: 38464 samples
Training acc over epoch: 0.8919
Validation acc: 0.8996
Time taken: 1.34s

Much faster, isn't it?

Low-level handling of losses tracked by the model

Layers & models recursively track any losses created during the forward pass by layers that call self.add_loss(value). The resulting list of scalar loss values are available via the property model.losses at the end of the forward pass.

If you want to be using these loss components, you should sum them and add them to the main loss in your training step.

Consider this layer, that creates an activity regularization loss:

@keras.saving.register_keras_serializable()
class ActivityRegularizationLayer(layers.Layer):
    def call(self, inputs):
        self.add_loss(1e-2 * tf.reduce_sum(inputs))
        return inputs

Let's build a really simple model that uses it:

inputs = keras.Input(shape=(784,), name="digits")
x = layers.Dense(64, activation="relu")(inputs)
# Insert activity regularization as a layer
x = ActivityRegularizationLayer()(x)
x = layers.Dense(64, activation="relu")(x)
outputs = layers.Dense(10, name="predictions")(x)

model = keras.Model(inputs=inputs, outputs=outputs)

Here's what our training step should look like now:

@tf.function
def train_step(x, y):
    with tf.GradientTape() as tape:
        logits = model(x, training=True)
        loss_value = loss_fn(y, logits)
        # Add any extra losses created during the forward pass.
        loss_value += sum(model.losses)
    grads = tape.gradient(loss_value, model.trainable_weights)
    optimizer.apply_gradients(zip(grads, model.trainable_weights))
    train_acc_metric.update_state(y, logits)
    return loss_value

Summary

Now you know everything there is to know about using built-in training loops and writing your own from scratch.

To conclude, here's a simple end-to-end example that ties together everything you've learned in this guide: a DCGAN trained on MNIST digits.

End-to-end example: a GAN training loop from scratch

You may be familiar with Generative Adversarial Networks (GANs). GANs can generate new images that look almost real, by learning the latent distribution of a training dataset of images (the "latent space" of the images).

A GAN is made of two parts: a "generator" model that maps points in the latent space to points in image space, a "discriminator" model, a classifier that can tell the difference between real images (from the training dataset) and fake images (the output of the generator network).

A GAN training loop looks like this:

1) Train the discriminator. - Sample a batch of random points in the latent space. - Turn the points into fake images via the "generator" model. - Get a batch of real images and combine them with the generated images. - Train the "discriminator" model to classify generated vs. real images.

2) Train the generator. - Sample random points in the latent space. - Turn the points into fake images via the "generator" network. - Get a batch of real images and combine them with the generated images. - Train the "generator" model to "fool" the discriminator and classify the fake images as real.

For a much more detailed overview of how GANs works, see Deep Learning with Python.

Let's implement this training loop. First, create the discriminator meant to classify fake vs real digits:

discriminator = keras.Sequential(
    [
        keras.Input(shape=(28, 28, 1)),
        layers.Conv2D(64, (3, 3), strides=(2, 2), padding="same"),
        layers.LeakyReLU(alpha=0.2),
        layers.Conv2D(128, (3, 3), strides=(2, 2), padding="same"),
        layers.LeakyReLU(alpha=0.2),
        layers.GlobalMaxPooling2D(),
        layers.Dense(1),
    ],
    name="discriminator",
)
discriminator.summary()

Model: "discriminator"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 conv2d (Conv2D)             (None, 14, 14, 64)        640       
                                                                 
 leaky_re_lu (LeakyReLU)     (None, 14, 14, 64)        0         
                                                                 
 conv2d_1 (Conv2D)           (None, 7, 7, 128)         73856     
                                                                 
 leaky_re_lu_1 (LeakyReLU)   (None, 7, 7, 128)         0         
                                                                 
 global_max_pooling2d (Glob  (None, 128)               0         
 alMaxPooling2D)                                                 
                                                                 
 dense_4 (Dense)             (None, 1)                 129       
                                                                 
=================================================================
Total params: 74625 (291.50 KB)
Trainable params: 74625 (291.50 KB)
Non-trainable params: 0 (0.00 Byte)
_________________________________________________________________

Then let's create a generator network, that turns latent vectors into outputs of shape (28, 28, 1) (representing MNIST digits):

latent_dim = 128

generator = keras.Sequential(
    [
        keras.Input(shape=(latent_dim,)),
        # We want to generate 128 coefficients to reshape into a 7x7x128 map
        layers.Dense(7 * 7 * 128),
        layers.LeakyReLU(alpha=0.2),
        layers.Reshape((7, 7, 128)),
        layers.Conv2DTranspose(128, (4, 4), strides=(2, 2), padding="same"),
        layers.LeakyReLU(alpha=0.2),
        layers.Conv2DTranspose(128, (4, 4), strides=(2, 2), padding="same"),
        layers.LeakyReLU(alpha=0.2),
        layers.Conv2D(1, (7, 7), padding="same", activation="sigmoid"),
    ],
    name="generator",
)

Here's the key bit: the training loop. As you can see it is quite straightforward. The training step function only takes 17 lines.

# Instantiate one optimizer for the discriminator and another for the generator.
d_optimizer = keras.optimizers.Adam(learning_rate=0.0003)
g_optimizer = keras.optimizers.Adam(learning_rate=0.0004)

# Instantiate a loss function.
loss_fn = keras.losses.BinaryCrossentropy(from_logits=True)


@tf.function
def train_step(real_images):
    # Sample random points in the latent space
    random_latent_vectors = tf.random.normal(shape=(batch_size, latent_dim))
    # Decode them to fake images
    generated_images = generator(random_latent_vectors)
    # Combine them with real images
    combined_images = tf.concat([generated_images, real_images], axis=0)

    # Assemble labels discriminating real from fake images
    labels = tf.concat(
        [tf.ones((batch_size, 1)), tf.zeros((real_images.shape[0], 1))], axis=0
    )
    # Add random noise to the labels - important trick!
    labels += 0.05 * tf.random.uniform(labels.shape)

    # Train the discriminator
    with tf.GradientTape() as tape:
        predictions = discriminator(combined_images)
        d_loss = loss_fn(labels, predictions)
    grads = tape.gradient(d_loss, discriminator.trainable_weights)
    d_optimizer.apply_gradients(zip(grads, discriminator.trainable_weights))

    # Sample random points in the latent space
    random_latent_vectors = tf.random.normal(shape=(batch_size, latent_dim))
    # Assemble labels that say "all real images"
    misleading_labels = tf.zeros((batch_size, 1))

    # Train the generator (note that we should *not* update the weights
    # of the discriminator)!
    with tf.GradientTape() as tape:
        predictions = discriminator(generator(random_latent_vectors))
        g_loss = loss_fn(misleading_labels, predictions)
    grads = tape.gradient(g_loss, generator.trainable_weights)
    g_optimizer.apply_gradients(zip(grads, generator.trainable_weights))
    return d_loss, g_loss, generated_images

Let's train our GAN, by repeatedly calling train_step on batches of images.

Since our discriminator and generator are convnets, you're going to want to run this code on a GPU.

import os

# Prepare the dataset. We use both the training & test MNIST digits.
batch_size = 64
(x_train, _), (x_test, _) = keras.datasets.mnist.load_data()
all_digits = np.concatenate([x_train, x_test])
all_digits = all_digits.astype("float32") / 255.0
all_digits = np.reshape(all_digits, (-1, 28, 28, 1))
dataset = tf.data.Dataset.from_tensor_slices(all_digits)
dataset = dataset.shuffle(buffer_size=1024).batch(batch_size)

epochs = 1  # In practice you need at least 20 epochs to generate nice digits.
save_dir = "./"

for epoch in range(epochs):
    print("\nStart epoch", epoch)

    for step, real_images in enumerate(dataset):
        # Train the discriminator & generator on one batch of real images.
        d_loss, g_loss, generated_images = train_step(real_images)

        # Logging.
        if step % 200 == 0:
            # Print metrics
            print("discriminator loss at step %d: %.2f" % (step, d_loss))
            print("adversarial loss at step %d: %.2f" % (step, g_loss))

            # Save one generated image
            img = keras.utils.array_to_img(generated_images[0] * 255.0, scale=False)
            img.save(os.path.join(save_dir, "generated_img" + str(step) + ".png"))

        # To limit execution time we stop after 10 steps.
        # Remove the lines below to actually train the model!
        if step > 10:
            break

Start epoch 0
discriminator loss at step 0: 0.72
adversarial loss at step 0: 0.72

That's it! You'll get nice-looking fake MNIST digits after just ~30s of training on the Colab GPU.