I'm trying to create a wrapper for my Tensorflow code so that I can invoke it from a C++ program. Here's a toy example code written in Cython.
# Saved as ufa.pyx
import tensorflow as tf
import numpy as np
def execute_tf():
print("Here")
return 0.0
#x = tf.placeholder(tf.float32, [1, 2])
#w = tf.constant([0.2, 0.3], tf.float32)
#b = tf.constant( -0.1, tf.float32)
#linearModel = tf.matmul(x, tf.reshape(w, shape=[2, 1])) + b
#session = tf.Session()
#result = session.run(linearModel, feed_dict={x: np.reshape([-0.1, -0.4], [1, 2])} )
#return np.reshape(result, [])
cdef public float executeTF():
return execute_tf()
As you can see, most of the lines of execute_tf are commented out. Here's the driver code written in C++
// Saved as driver.cpp
#include "Python.h"
#include "ufa.h"
#include <iostream>
using namespace std;
int main(int argc, char *argv[])
{
Py_Initialize();
PyInit_ufa();
cout << "Hello world" << endl;
executeTF();
Py_Finalize();
return 0;
}
I compile it as follows
$ cython --cplus ufa.pyx
$ g++ ufa.cpp driver.cpp -L$(python3.6-config --cflags) -I$(python3.6-config --ldflags) -std=c++11 -o test.opt
If I compile this while commenting out import tensorflow as tf, the program runs fine. However, if I import tensorflow, I get the following error
Hello world
NameError: name 'execute_tf' is not defined
Exception ignored in: 'ufa.executeTF'
NameError: name 'execute_tf' is not defined
Any idea why it is happening? Here's some more details of the system configuration:
- Python: 3.6.10
- Cython: 0.26
- GCC: 8.3.1
- Tensorflow: 1.15.0
- OS: CentOS
