____tz_zs
tensorflow 提供tf.train.Saver类来保存和加载神经网络模型,saver.save()保存模型,saver.restore加载模型
源码saver.py中定义的保存方法
def save(self,
sess,
save_path,
global_step=None,
latest_filename=None,
meta_graph_suffix="meta",
write_meta_graph=True,
write_state=True):
sess:用于保存变量的会话
save_path:保存的路径名
global_step:如果提供了全局步数,则在保存文件的末尾加上训练轮数
latest_filename:协议缓冲区文件的可选名称
meta_graph_suffix: `MetaGraphDef` 文件的后缀,默认为 'meta'
write_meta_graph:标示是否写入元图文件
write_state:标示是否写 `CheckpointStateProto`文件
code1:使用的小Demo
# -*- coding: utf-8 -*-
"""
@author: tz_zs
持久化实现 模型的保存和加载
"""
import tensorflow as tf
v1 = tf.Variable(tf.constant(1.0, shape=[1]), name="v1")
v2 = tf.Variable(tf.constant(2.0, shape=[1]), name="v2")
result1 = v1 + v2
init_op = tf.global_variables_initializer()
saver = tf.train.Saver()
with tf.Session() as sess:
sess.run(init_op)
# 保存模型到/path/to/model/model.ckpt文件
saver.save(sess, "/path/to/model/model.ckpt")
'''
保存模型到了/path/to/model/model.ckpt文件中。
此时会生成三个文件:
model.ckpt.meta
model.ckpt
checkpoint
'''
# 加载模型
saver2 = tf.train.import_meta_graph("/path/to/model/model.ckpt.meta")
with tf.Session() as sess:
saver2.restore(sess, "/path/to/model/model.ckpt")
print(sess.run(tf.get_default_graph().get_tensor_by_name("add:0"))) # [ 3.]
code2:使用 convert_variables_to_constants 这个函数,可以将计算图中的变量及其取值通过常量的方式保存,写入单个文件
# -*- coding: utf-8 -*-
"""
@author: tz_zs
单文件 常量的方式保存
convert_variables_to_constants
"""
import tensorflow as tf
from tensorflow.python.framework import graph_util
from tensorflow.python.platform import gfile
v1 = tf.Variable(1, dtype=tf.float32, name="v1")
v2 = tf.Variable(2, dtype=tf.float32, name="v2")
result = v1 + v2
with tf.Session() as sess:
tf.global_variables_initializer().run()
graph_def = tf.get_default_graph().as_graph_def()
print("#####" * 5 + " 得到当前默认的计算图graph_def " + "#####" * 5)
print(graph_def)
print("#####" * 5 + " add节点相关的output_graph_def " + "#####" * 5)
output_graph_def = graph_util.convert_variables_to_constants(sess, graph_def, ["add"])
print(output_graph_def)
with tf.gfile.GFile("/path/to/model/combined_model.pb", "wb") as f:
f.write(output_graph_def.SerializeToString())
print("#####" * 5 + " 加载模型 " + "#####" * 5)
with tf.Session() as sess:
model_filename = "/path/to/model/combined_model.pb"
with gfile.FastGFile(model_filename, 'rb') as f:
graph_def2 = tf.GraphDef()
graph_def2.ParseFromString(f.read())
print(graph_def2)
# 可以通过张量的名称取某个张量的值
result = tf.import_graph_def(graph_def2, return_elements=["add:0"])
result_v1 = tf.import_graph_def(graph_def2, return_elements=["v1:0"])
print(result)
print(sess.run(result))
print(result_v1)
print(sess.run(result_v1))
运行的输出结果:
######################### 得到当前默认的计算图graph_def #########################
node {
name: "v1/initial_value"
op: "Const"
attr {
key: "dtype"
value {
type: DT_FLOAT
}
}
attr {
key: "value"
value {
tensor {
dtype: DT_FLOAT
tensor_shape {
}
float_val: 1.0
}
}
}
}
node {
name: "v1"
op: "VariableV2"
attr {
key: "container"
value {
s: ""
}
}
attr {
key: "dtype"
value {
type: DT_FLOAT
}
}
attr {
key: "shape"
value {
shape {
}
}
}
attr {
key: "shared_name"
value {
s: ""
}
}
}
node {
name: "v1/Assign"
op: "Assign"
input: "v1"
input: "v1/initial_value"
attr {
key: "T"
value {
type: DT_FLOAT
}
}
attr {
key: "_class"
value {
list {
s: "loc:@v1"
}
}
}
attr {
key: "use_locking"
value {
b: true
}
}
attr {
key: "validate_shape"
value {
b: true
}
}
}
node {
name: "v1/read"
op: "Identity"
input: "v1"
attr {
key: "T"
value {
type: DT_FLOAT
}
}
attr {
key: "_class"
value {
list {
s: "loc:@v1"
}
}
}
}
node {
name: "v2/initial_value"
op: "Const"
attr {
key: "dtype"
value {
type: DT_FLOAT
}
}
attr {
key: "value"
value {
tensor {
dtype: DT_FLOAT
tensor_shape {
}
float_val: 2.0
}
}
}
}
node {
name: "v2"
op: "VariableV2"
attr {
key: "container"
value {
s: ""
}
}
attr {
key: "dtype"
value {
type: DT_FLOAT
}
}
attr {
key: "shape"
value {
shape {
}
}
}
attr {
key: "shared_name"
value {
s: ""
}
}
}
node {
name: "v2/Assign"
op: "Assign"
input: "v2"
input: "v2/initial_value"
attr {
key: "T"
value {
type: DT_FLOAT
}
}
attr {
key: "_class"
value {
list {
s: "loc:@v2"
}
}
}
attr {
key: "use_locking"
value {
b: true
}
}
attr {
key: "validate_shape"
value {
b: true
}
}
}
node {
name: "v2/read"
op: "Identity"
input: "v2"
attr {
key: "T"
value {
type: DT_FLOAT
}
}
attr {
key: "_class"
value {
list {
s: "loc:@v2"
}
}
}
}
node {
name: "add"
op: "Add"
input: "v1/read"
input: "v2/read"
attr {
key: "T"
value {
type: DT_FLOAT
}
}
}
node {
name: "init"
op: "NoOp"
input: "^v1/Assign"
input: "^v2/Assign"
}
versions {
producer: 22
}
######################### add节点相关的output_graph_def #########################
Converted 2 variables to const ops.
node {
name: "v1"
op: "Const"
attr {
key: "dtype"
value {
type: DT_FLOAT
}
}
attr {
key: "value"
value {
tensor {
dtype: DT_FLOAT
tensor_shape {
}
float_val: 1.0
}
}
}
}
node {
name: "v1/read"
op: "Identity"
input: "v1"
attr {
key: "T"
value {
type: DT_FLOAT
}
}
attr {
key: "_class"
value {
list {
s: "loc:@v1"
}
}
}
}
node {
name: "v2"
op: "Const"
attr {
key: "dtype"
value {
type: DT_FLOAT
}
}
attr {
key: "value"
value {
tensor {
dtype: DT_FLOAT
tensor_shape {
}
float_val: 2.0
}
}
}
}
node {
name: "v2/read"
op: "Identity"
input: "v2"
attr {
key: "T"
value {
type: DT_FLOAT
}
}
attr {
key: "_class"
value {
list {
s: "loc:@v2"
}
}
}
}
node {
name: "add"
op: "Add"
input: "v1/read"
input: "v2/read"
attr {
key: "T"
value {
type: DT_FLOAT
}
}
}
library {
}
######################### 加载模型 #########################
node {
name: "v1"
op: "Const"
attr {
key: "dtype"
value {
type: DT_FLOAT
}
}
attr {
key: "value"
value {
tensor {
dtype: DT_FLOAT
tensor_shape {
}
float_val: 1.0
}
}
}
}
node {
name: "v1/read"
op: "Identity"
input: "v1"
attr {
key: "T"
value {
type: DT_FLOAT
}
}
attr {
key: "_class"
value {
list {
s: "loc:@v1"
}
}
}
}
node {
name: "v2"
op: "Const"
attr {
key: "dtype"
value {
type: DT_FLOAT
}
}
attr {
key: "value"
value {
tensor {
dtype: DT_FLOAT
tensor_shape {
}
float_val: 2.0
}
}
}
}
node {
name: "v2/read"
op: "Identity"
input: "v2"
attr {
key: "T"
value {
type: DT_FLOAT
}
}
attr {
key: "_class"
value {
list {
s: "loc:@v2"
}
}
}
}
node {
name: "add"
op: "Add"
input: "v1/read"
input: "v2/read"
attr {
key: "T"
value {
type: DT_FLOAT
}
}
}
library {
}
[<tf.Tensor 'import/add:0' shape=() dtype=float32>]
[3.0]
[<tf.Tensor 'import_1/v1:0' shape=() dtype=float32>]
[1.0]
Process finished with exit code 0