查看tensorflow pb模型文件的节点信息
2024-08-29 02:14:50
查看tensorflow pb模型文件的节点信息：
import tensorflow as tf
with tf.Session() as sess:
with open('./quantized_model.pb', 'rb') as f:
graph_def = tf.GraphDef()
graph_def.ParseFromString(f.read())
print graph_def
效果：
# ...
node {
name: "FullyConnected/BiasAdd"
op: "BiasAdd"
input: "FullyConnected/MatMul"
input: "FullyConnected/b/read"
attr {
key: "T"
value {
type: DT_FLOAT
}
}
attr {
key: "data_format"
value {
s: "NHWC"
}
}
}
node {
name: "FullyConnected/Softmax"
op: "Softmax"
input: "FullyConnected/BiasAdd"
attr {
key: "T"
value {
type: DT_FLOAT
}
}
}
library {
}
参考：https://tang.su/2017/01/export-TensorFlow-network/
https://github.com/tensorflow/tensorflow/issues/15689
一些核心代码：
import tensorflow as tf
with tf.Session() as sess:
with open('./graph.pb', 'rb') as f:
graph_def = tf.GraphDef()
graph_def.ParseFromString(f.read())
print graph_def
output = tf.import_graph_def(graph_def, return_elements=['out:0'])
print(sess.run(output))
This is part of my Tensorflow frozen graph, I have named the input and output nodes.
>>> g.ParseFromString(open('frozen_graph.pb','rb').read())
>>> g
node {
name: "input"
op: "Placeholder"
attr {
key: "dtype"
value {
type: DT_FLOAT
}
}
attr {
key: "shape"
value {
shape {
dim {
size: -1
}
dim {
size: 68
}
}
}
}
}
...
node {
name: "output"
op: "Softmax"
input: "add"
attr {
key: "T"
value {
type: DT_FLOAT
}
}
}
I ran this model by the following code
(CELL is name of directory where my file is located)
final String MODEL_FILE = "file:///android_asset/" + CELL + "/optimized_graph.pb" ;
final String INPUT_NODE = "input" ;
final String OUTPUT_NODE = "output" ;
final int[] INPUT_SIZE = {1,68} ;
float[] RESULT = new float[8];
inferenceInterface = new TensorFlowInferenceInterface();
inferenceInterface.initializeTensorFlow(getAssets(),MODEL_FILE) ;
inferenceInterface.fillNodeFloat(INPUT_NODE,INPUT_SIZE,input);
and finally
inferenceInterface.readNodeFloat(OUTPUT_NODE,RESULT);
最新文章
- lucene 索引 demo
- Html5新标签解释及用法
- php关于static关键字
- Nginx工作原理和优化、漏洞
- Nginx下Redmine2.6配置
- How to manage and balance “Huge Data Load” for Big Kafka Clusters---reference
- IE8下ckeditor无法正常使用，提示"例外被抛出且未被接住"的解决办法
- vim乱码问题
- 抽象类 abstract 和 接口 interface 类的区别
- MERGE语法详解
- 在用jQuery时遇到的小问题
- 面试必问---HashMap原理分析
- 剑指offer-面试题7：俩个栈实现队列（java）
- PowerDesigner 提示 Existence of index、key、reference错误
- @Autowired 基本介绍，有待丰富
- MapReduce实例学习
- mysql主从配置，读写分离
- 5. 多重背包问题 II 【用二进制优化】
- MFQ
- 超简单，centos7安装docker