Guia de início rápido do TensorFlow 2 para especialistas

Este é um arquivo de notebook do Google Colaboratory . Os programas Python são executados diretamente no navegador, uma ótima maneira de aprender e usar o TensorFlow. Para seguir este tutorial, execute o notebook no Google Colab clicando no botão na parte superior desta página.

  1. No Colab, conecte-se a um tempo de execução do Python: no canto superior direito da barra de menus, selecione CONNECT .
  2. Execute todas as células de código do notebook: selecione Runtime > Run all .

Baixe e instale o TensorFlow 2. Importe o TensorFlow para o seu programa:

Importe o TensorFlow para o seu programa:

import tensorflow as tf
print("TensorFlow version:", tf.__version__)

from tensorflow.keras.layers import Dense, Flatten, Conv2D
from tensorflow.keras import Model
TensorFlow version: 2.8.0-rc1

Carregue e prepare o conjunto de dados MNIST .

mnist = tf.keras.datasets.mnist

(x_train, y_train), (x_test, y_test) = mnist.load_data()
x_train
, x_test = x_train / 255.0, x_test / 255.0

# Add a channels dimension
x_train
= x_train[..., tf.newaxis].astype("float32")
x_test
= x_test[..., tf.newaxis].astype("float32")

Use tf.data para agrupar e embaralhar o conjunto de dados:

train_ds = tf.data.Dataset.from_tensor_slices(
   
(x_train, y_train)).shuffle(10000).batch(32)

test_ds
= tf.data.Dataset.from_tensor_slices((x_test, y_test)).batch(32)

Crie o modelo tf.keras usando a API de subclassificação do modelo Keras :

class MyModel(Model):
 
def __init__(self):
   
super(MyModel, self).__init__()
   
self.conv1 = Conv2D(32, 3, activation='relu')
   
self.flatten = Flatten()
   
self.d1 = Dense(128, activation='relu')
   
self.d2 = Dense(10)

 
def call(self, x):
    x
= self.conv1(x)
    x
= self.flatten(x)
    x
= self.d1(x)
   
return self.d2(x)

# Create an instance of the model
model
= MyModel()

Escolha um otimizador e uma função de perda para treinamento:

loss_object = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True)

optimizer
= tf.keras.optimizers.Adam()

Selecione métricas para medir a perda e a precisão do modelo. Essas métricas acumulam os valores ao longo das épocas e, em seguida, imprimem o resultado geral.

train_loss = tf.keras.metrics.Mean(name='train_loss')
train_accuracy
= tf.keras.metrics.SparseCategoricalAccuracy(name='train_accuracy')

test_loss
= tf.keras.metrics.Mean(name='test_loss')
test_accuracy
= tf.keras.metrics.SparseCategoricalAccuracy(name='test_accuracy')

Use tf.GradientTape para treinar o modelo:

@tf.function
def train_step(images, labels):
 
with tf.GradientTape() as tape:
   
# training=True is only needed if there are layers with different
   
# behavior during training versus inference (e.g. Dropout).
    predictions
= model(images, training=True)
    loss
= loss_object(labels, predictions)
  gradients
= tape.gradient(loss, model.trainable_variables)
  optimizer
.apply_gradients(zip(gradients, model.trainable_variables))

  train_loss
(loss)
  train_accuracy
(labels, predictions)

Teste o modelo:

@tf.function
def test_step(images, labels):
 
# training=False is only needed if there are layers with different
 
# behavior during training versus inference (e.g. Dropout).
  predictions
= model(images, training=False)
  t_loss
= loss_object(labels, predictions)

  test_loss
(t_loss)
  test_accuracy
(labels, predictions)
EPOCHS = 5

for epoch in range(EPOCHS):
 
# Reset the metrics at the start of the next epoch
  train_loss
.reset_states()
  train_accuracy
.reset_states()
  test_loss
.reset_states()
  test_accuracy
.reset_states()

 
for images, labels in train_ds:
    train_step
(images, labels)

 
for test_images, test_labels in test_ds:
    test_step
(test_images, test_labels)

 
print(
    f
'Epoch {epoch + 1}, '
    f
'Loss: {train_loss.result()}, '
    f
'Accuracy: {train_accuracy.result() * 100}, '
    f
'Test Loss: {test_loss.result()}, '
    f
'Test Accuracy: {test_accuracy.result() * 100}'
 
)
Epoch 1, Loss: 0.13306719064712524, Accuracy: 96.03833770751953, Test Loss: 0.0717063844203949, Test Accuracy: 97.68999481201172
Epoch 2, Loss: 0.04493752866983414, Accuracy: 98.61833190917969, Test Loss: 0.058997876942157745, Test Accuracy: 98.18000030517578
Epoch 3, Loss: 0.023821160197257996, Accuracy: 99.22000122070312, Test Loss: 0.0560370571911335, Test Accuracy: 98.30999755859375
Epoch 4, Loss: 0.014193248935043812, Accuracy: 99.50666809082031, Test Loss: 0.06797954440116882, Test Accuracy: 98.29999542236328
Epoch 5, Loss: 0.010457769967615604, Accuracy: 99.63666534423828, Test Loss: 0.08524733036756516, Test Accuracy: 97.83999633789062

O classificador de imagem agora é treinado com precisão de ~ 98% neste conjunto de dados. Para saber mais, leia os tutoriais do TensorFlow .