tensorflow2的模型保存与加载

模型的保存与加载

save/load weights

轻量级方式，只保存网络的参数，不管其他的状态，这种模式适合自己对代码有个清晰的认识。我们在training时如果中断了，可以通过这个方式保存最近更新的参数。

例：创建一个网络结构，首先training、test，将参数保存进weights.ckpt，删除网络结构，在创建一个原来一样的网络结构，导入保存的参数，进行test，会发现loss和accuracy与删除前的值相近（网络结构中有很多其他的因子会影响最后的结果，所以不一定完全相同）

#save the weights
network.save_weights('./checkpoints/my_checkpoint')
# Restore the weights
model = create_model()
model.load_weights('./checkpoints/my_checkpoint')
loss,acc = model.evaluate(test_images, test_labels)
print("Restored model, accuracy:{:5.2f}".format(100*acc))

完整代码实例：

import  os
os.environ['TF_CPP_MIN_LOG_LEVEL']='2'

import  tensorflow as tf
from    tensorflow.keras import datasets, layers, optimizers, Sequential, metrics


def preprocess(x, y):
    """
    x is a simple image, not a batch
    """
    x = tf.cast(x, dtype=tf.float32) / 255.
    x = tf.reshape(x, [28*28])
    y = tf.cast(y, dtype=tf.int32)
    y = tf.one_hot(y, depth=10)
    return x,y


batchsz = 128
(x, y), (x_val, y_val) = datasets.mnist.load_data()
print('datasets:', x.shape, y.shape, x.min(), x.max())



db = tf.data.Dataset.from_tensor_slices((x,y))
db = db.map(preprocess).shuffle(60000).batch(batchsz)
ds_val = tf.data.Dataset.from_tensor_slices((x_val, y_val))
ds_val = ds_val.map(preprocess).batch(batchsz) 

sample = next(iter(db))
print(sample[0].shape, sample[1].shape)


network = Sequential([layers.Dense(256, activation='relu'),
                     layers.Dense(128, activation='relu'),
                     layers.Dense(64, activation='relu'),
                     layers.Dense(32, activation='relu'),
                     layers.Dense(10)])
network.build(input_shape=(None, 28*28))
network.summary()

network.compile(optimizer=optimizers.Adam(lr=0.01),
		loss=tf.losses.CategoricalCrossentropy(from_logits=True),
		metrics=['accuracy']
	)

network.fit(db, epochs=3, validation_data=ds_val, validation_freq=2)
 
network.evaluate(ds_val)

network.save_weights('weights.ckpt')
print('saved weights.')
del network
#training、test完成后，将参数保存进weights.ckpt，然后删除前面的网络

#创建一个原来一样的新的网络
network = Sequential([layers.Dense(256, activation='relu'),
                     layers.Dense(128, activation='relu'),
                     layers.Dense(64, activation='relu'),
                     layers.Dense(32, activation='relu'),
                     layers.Dense(10)])
network.compile(optimizer=optimizers.Adam(lr=0.01),
		loss=tf.losses.CategoricalCrossentropy(from_logits=True),
		metrics=['accuracy']
	)
network.load_weights('weights.ckpt')#导入上一次保存的参数
print('loaded weights!')
network.evaluate(ds_val)  #测试网络

save/load entire model

最简单粗暴的方式，将所有的状态都保存起来，可以完美的进行恢复。不需要重新创建网络结构，测试的结果也与上一次完全一样。

network.save('model.h5')
print('saved total model.')
del network

print('loaded model from file.')
network = tf.keras.models.load_model('model.h5', compile=False) 
network.evaluate(ds_val) 

完整代码实例：

import  os
os.environ['TF_CPP_MIN_LOG_LEVEL']='2'

import  tensorflow as tf
from    tensorflow.keras import datasets, layers, optimizers, Sequential, metrics


def preprocess(x, y):
    """
    x is a simple image, not a batch
    """
    x = tf.cast(x, dtype=tf.float32) / 255.
    x = tf.reshape(x, [28*28])
    y = tf.cast(y, dtype=tf.int32)
    y = tf.one_hot(y, depth=10)
    return x,y


batchsz = 128
(x, y), (x_val, y_val) = datasets.mnist.load_data()
print('datasets:', x.shape, y.shape, x.min(), x.max())



db = tf.data.Dataset.from_tensor_slices((x,y))
db = db.map(preprocess).shuffle(60000).batch(batchsz)
ds_val = tf.data.Dataset.from_tensor_slices((x_val, y_val))
ds_val = ds_val.map(preprocess).batch(batchsz) 

sample = next(iter(db))
print(sample[0].shape, sample[1].shape)


network = Sequential([layers.Dense(256, activation='relu'),
                     layers.Dense(128, activation='relu'),
                     layers.Dense(64, activation='relu'),
                     layers.Dense(32, activation='relu'),
                     layers.Dense(10)])
network.build(input_shape=(None, 28*28))
network.summary()




network.compile(optimizer=optimizers.Adam(lr=0.01),
		loss=tf.losses.CategoricalCrossentropy(from_logits=True),
		metrics=['accuracy']
	)

network.fit(db, epochs=3, validation_data=ds_val, validation_freq=2)
 
network.evaluate(ds_val)  #测试

network.save('model.h5')  #保存
print('saved total model.')
del network  #删除网络结构

print('loaded model from file.')
network = tf.keras.models.load_model('model.h5', compile=False)   #导入保存的数据
network.compile(optimizer=optimizers.Adam(lr=0.01),
        loss=tf.losses.CategoricalCrossentropy(from_logits=True),
        metrics=['accuracy']
    )
x_val = tf.cast(x_val, dtype=tf.float32) / 255.
x_val = tf.reshape(x_val, [-1, 28*28])
y_val = tf.cast(y_val, dtype=tf.int32)
y_val = tf.one_hot(y_val, depth=10)
ds_val = tf.data.Dataset.from_tensor_slices((x_val, y_val)).batch(128)
network.evaluate(ds_val)  #测试

saved_model
模型的一种保存格式，跟pytorch的ONNX对应，也就是说当训练的一个模型交给工厂的生产环境的时候，可以把这个模型直接交给用户来部署，而不需要给一个源代码或相关的信息，这个模型就包含的所有的这样一个信息。比如，你通过python写的源文件，你可以用c++解析和读取这个工作。详细说明看这里

tf.saved_model.save(m, '/tmp/saved_model/')
imported = tf.saved_model.load(path)
f = imported.signatures["serving_default"]
print(f(x = tf.ones[1, 28, 28, 3]))