[tf1] fatal error: 'tensorflow/core/framework/op_kernel.h' file not found about tf-approximate HOT 6 CLOSED

Hi John,
this version was written for TensorFlow 1 (1.14, but it is working even in version 1.15). If you need to use TF 2.0, please use GPU version (distributed as a Singularity image).

tensorflow                1.15.0          gpu_py37h0f0df58_0
tensorflow-base           1.15.0          gpu_py37h9dcbed7_0
tensorflow-estimator      1.15.1             pyh2649769_0

Did it help?

from tf-approximate.

thanks, I'll try again.

from tf-approximate.

Hi, My enviroment is ubuntu16.04 with anaconda3(python3.7, tf1.15).
I've found the op_kernel.h in my enviroment. The path is /anaconda3/envs/tf1_15/lib/python3.7/site-packages/tensorflow_core/include/tensorflow/core/framework/op_kernel.h. Then I copy all the file in tf1/axqconv into the path, /anaconda3/envs/tf1_15/lib/python3.7/site-packages/tensorflow_core/include/, and run the Makefile.But there is another error as follow:
(tf1_15) zhouhang@ubuntu:/anaconda3/envs/tf1_15/lib/python3.7/site-packages/tensorflow_core/include$ make
g++ -o tables tables.cc axmult.o
In file included from /usr/include/c++/5/cstdint:35:0,
from approximate_selector.h:2,
from tables.cc:7:
/usr/include/c++/5/bits/c++0x_warning.h:32:2: error: #error This file requires compiler and library support for the ISO C++ 2011 standard. This support must be enabled with the -std=c++11 or -std=gnu++11 compiler options.
#error This file requires compiler and library support
^
Makefile:27: recipe for target 'tables' failed
make: *** [tables] Error 1
Perhaps, the g++'s version cause the error. In my machine, gcc version is 5.4.0 20160609 (Ubuntu 5.4.0-6ubuntu116.04.12) .

from tf-approximate.

4

I modified two lines of code in the Makefile by adding the -std=c++11 as follow:
tune: tune.o axmult.o
g++ -std=c++11 -o $@ $^

tables: tables.cc axmult.o approximate_selector.h
g++ -std=c++11 -o $@ $(filter-out %.h, $^)
Then run the Makefile, and axqconv.so is generated successfully. Does it result in bad outcome?

from tf-approximate.

You can try to run

make clean

To delete file .tfconfig, where the configuration for TensorFlow extension building is cached. If you changed your TF version using different Anaconda environment, the build arguments may change.

from tf-approximate.

thank you.

from tf-approximate.