1 保存序列模型或函数式模型
In [1]:
# 构建一个简单的模型并训练
from __future__ import absolute_import, division, print_function
import tensorflow as tf
tf.keras.backend.clear_session()
from tensorflow import keras
from tensorflow.keras import layers
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, activation='softmax', name='predictions')(x)
model = keras.Model(inputs=inputs, outputs=outputs, name='3_layer_mlp')
model.summary()
(x_train, y_train), (x_test, y_test) = keras.datasets.mnist.load_data()
x_train = x_train.reshape(60000, 784).astype('float32') / 255
x_test = x_test.reshape(10000, 784).astype('float32') / 255
model.compile(loss='sparse_categorical_crossentropy',
optimizer=keras.optimizers.RMSprop())
history = model.fit(x_train, y_train,
batch_size=64,
epochs=1)
predictions = model.predict(x_test)
1.1 保存整个模型
可以对整个模型进行保存,其保持的内容包括:
- 该模型的架构
- 模型的权重(在训练期间学到的)
- 模型的训练配置(传递给编译的)
- 优化器及其状态(这使您可以从中断的地方重新启动训练)
In [2]:
import numpy as np
# 模型保存
model.save('the_save_model.h5')
# 导入模型
new_model = keras.models.load_model('the_save_model.h5')
new_prediction = new_model.predict(x_test)
np.testing.assert_allclose(predictions, new_prediction, atol=1e-6) # 预测结果一样
1.2 导出为SavedModel文件
SavedModel是Tensorflow对象的独立序列化格式,支持使用Tensorflow Serving server来部署模型,支持其他语言读取。
In [4]:
# 导出为tf的SavedModel文件
model.save('save_model', save_format='tf')
# 从SavedModel文件中导入模型
new_model = keras.models.load_model('save_model')
new_prediction = new_model.predict(x_test)
np.testing.assert_allclose(predictions, new_prediction, atol=1e-6) # 预测结果一样
SaveModel创建的文件包含:
- 权重
- 网络图
1.3 仅保存网络结构
仅保持网络结构,这样导出的模型并未包含训练好的参数
In [5]:
# 获取网络结构配置
config = model.get_config()
reinitialized_model = keras.Model.from_config(config)
new_prediction = reinitialized_model.predict(x_test)
assert abs(np.sum(predictions-new_prediction)) >0
也可以使用json保存网络结构
In [6]:
# 将网络结构导出为json格式
json_config = model.to_json()
reinitialized_model = keras.models.model_from_json(json_config)
new_prediction = reinitialized_model.predict(x_test)
assert abs(np.sum(predictions-new_prediction)) >0
1.4 仅保存网络权重参数
In [7]:
# 获取网络权重
weights = model.get_weights()
# 对网络权重进行赋值
model.set_weights(weights)
可以把结构和参数保存结合起来
In [8]:
config = model.get_config()
weights = model.get_weights()
new_model = keras.Model.from_config(config) # config只能用keras.Model的这个api
new_model.set_weights(weights)
new_predictions = new_model.predict(x_test)
np.testing.assert_allclose(predictions, new_predictions, atol=1e-6)
1.5 完整的模型保存方法
In [10]:
# 导出网络结构和权重
json_config = model.to_json()
with open('model_config.json', 'w') as json_file:
json_file.write(json_config)
model.save_weights('path_to_my_weights.h5')
# 载入网络结构和权重
with open('model_config.json') as json_file:
json_config = json_file.read()
new_model = keras.models.model_from_json(json_config)
new_model.load_weights('path_to_my_weights.h5')
new_predictions = new_model.predict(x_test)
np.testing.assert_allclose(predictions, new_predictions, atol=1e-6)
In [11]:
# 当然也可以一步到位
model.save('path_to_my_model.h5')
del model
model = keras.models.load_model('path_to_my_model.h5')
1.6 权重保存格式
有.h5或.keras后缀时保存为keras HDF5格式文件,否则默认为TensorFlow Checkpoint格式文件。可以使用save_format显式确定。
In [12]:
model.save_weights('weight_tf_savedmodel')
model.save_weights('weight_tf_savedmodel.h5')
In [13]:
model.save_weights('weight_tf_savedmodel_tf', save_format='tf')
model.save_weights('weight_tf_savedmodel_h5', save_format='h5')
1.7 子类模型权重保存
子类模型的结构无法保存和序列化,只能保持参数
In [16]:
# 构建模型
class ThreeLayerMLP(keras.Model):
def __init__(self, name=None):
super(ThreeLayerMLP, self).__init__(name=name)
self.dense_1 = layers.Dense(64, activation='relu', name='dense_1')
self.dense_2 = layers.Dense(64, activation='relu', name='dense_2')
self.pred_layer = layers.Dense(10, activation='softmax', name='predictions')
def call(self, inputs):
x = self.dense_1(inputs)
x = self.dense_2(x)
return self.pred_layer(x)
def get_model():
return ThreeLayerMLP(name='3_layer_mlp')
model = get_model()
首先,无法保存从未使用过的子类模型。
这是因为需要在某些数据上调用子类模型才能创建其权重。
In [17]:
# 训练模型
(x_train, y_train), (x_test, y_test) = keras.datasets.mnist.load_data()
x_train = x_train.reshape(60000, 784).astype('float32') / 255
x_test = x_test.reshape(10000, 784).astype('float32') / 255
model.compile(loss='sparse_categorical_crossentropy',
optimizer=keras.optimizers.RMSprop())
history = model.fit(x_train, y_train,
batch_size=64,
epochs=1)
推荐的保存子类模型的方法是使用save_weights创建TensorFlow SavedModel检查点。
该检查点将包含与模型关联的所有变量的值:
- 图层的权重
- 优化器的状态
- 与有状态模型指标关联的任何变量
In [18]:
# 保持权重参数
model.save_weights('my_model_weights', save_format='tf')
# 输出结果,供后面对比
predictions = model.predict(x_test)
first_batch_loss = model.train_on_batch(x_train[:64], y_train[:64])
要还原模型,将需要访问创建模型对象的代码。
请注意,为了恢复优化器状态和任何有状态度量的状态,应该先编译模型(使用与以前完全相同的参数)。
In [19]:
# 读取保存的模型参数
new_model = get_model()
new_model.compile(loss='sparse_categorical_crossentropy',
optimizer=keras.optimizers.RMSprop())
#new_model.train_on_batch(x_train[:1], y_train[:1])
new_model.load_weights('my_model_weights')
new_predictions = new_model.predict(x_test)
np.testing.assert_allclose(predictions, new_predictions, atol=1e-6)
new_first_batch_loss = new_model.train_on_batch(x_train[:64], y_train[:64])
assert first_batch_loss == new_first_batch_loss
In [ ]:
本文来自博客园,作者:大码王,转载请注明原文链接:https://www.cnblogs.com/huanghanyu/