Keras 模块化 API 怎么用？2025 教程教你高效训练部署神经网络

from keras import layers, models

inputs = layers.Input(shape=(, , ))
x = layers.Conv2D(, (, ), activation='relu')(inputs)
x = layers.MaxPooling2D((, ))(x)
x = layers.Conv2D(, (, ), activation='relu')(x)
x = layers.MaxPooling2D((, ))(x)
x = layers.Flatten()(x)
x = layers.Dense(, activation='relu')(x)
outputs = layers.Dense(, activation='softmax')(x)

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

model.compile(optimizer='adam',
              loss='categorical_crossentropy',
              metrics=['accuracy'])

from torch.utils.data import DataLoader

train_loader = DataLoader(train_dataset, batch_size=, shuffle=True)
model.fit(train_loader, epochs=)

import keras.distribution as dist

strategy = dist.MultiDeviceStrategy(['device:GPU:0', 'device:GPU:1'])
with strategy.scope():
    model = build_model()

from kerastuner import RandomSearch

def build_model(hp):
    model = models.Sequential()
    model.add(layers.Input(shape=(, , )))
    for i in range(hp.Int('conv_layers', , )):
        model.add(layers.Conv2D(hp.Int(f'conv_{i}_filters', , , step=),
                               (, ), activation='relu'))
        model.add(layers.MaxPooling2D((, )))
    model.add(layers.Flatten())
    model.add(layers.Dense(hp.Int('dense_units', , , step=),
                          activation='relu'))
    model.add(layers.Dense(, activation='softmax'))
    model.compile(optimizer='adam',
                  loss='categorical_crossentropy',
                  metrics=['accuracy'])
    return model

tuner = RandomSearch(build_model,
                     objective='val_accuracy',
                     max_trials=)
tuner.search(train_images, train_labels, epochs=, validation_split=0.2)

test_loss, test_acc = model.evaluate(test_images, test_labels)
print(f'Test accuracy: {test_acc:.4f}')

predictions = model.predict(test_images)

model.save('my_model')

converter = tf.lite.TFLiteConverter.from_keras_model(model)
converter.optimizations = [tf.lite.Optimize.DEFAULT]
converter.target_spec.supported_types = [tf.float16]
tflite_model = converter.convert()

import keras.onnx

keras.onnx.save_model(model, 'my_model.onnx')

from keras.callbacks import EarlyStopping

early_stopping = EarlyStopping(monitor='val_loss',
                              patience=,
                              restore_best_weights=True)
model.fit(train_images, train_labels,
          epochs=,
          batch_size=,
          validation_split=0.2,
          callbacks=[early_stopping])

from keras.callbacks import ModelCheckpoint

model_checkpoint = ModelCheckpoint('best_model.h5',
                                  monitor='val_accuracy',
                                  save_best_only=True)

from keras.callbacks import TensorBoard

tensorboard = TensorBoard(log_dir='./logs')