Tensorflow C++ API and Caffe Compatibility issue related to Glog -
i trying write code in c++ using eclipse ide without bazel. project had perform hacks generate tensorflow c++ apis. tensorflow c++ seems working fine. can load model , perform predictions. have code uses caffe library else. these both seem work fine separately.
the problem
i want integrate both tensorflow , caffe in code. when include both libraries following issue:
in file included /usr/local/include/tf/tensorflow/core/platform/logging.h:25:0, /usr/local/include/tf/tensorflow/core/lib/core/status.h:25, /usr/local/include/tf/tensorflow/core/framework/variant.h:28, /usr/local/include/tf/tensorflow/core/framework/allocator.h:26, /usr/local/include/tf/tensorflow/core/framework/tensor.h:22, /usr/local/include/tf/tensorflow/cc/framework/ops.h:21, /usr/local/include/tf/tensorflow/cc/ops/const_op.h:19, /usr/local/include/tf/tensorflow/core/platform/default/logging.h:87:0: warning: "log" redefined [enabled default] #define log(severity) _tf_log_##severity ^ in file included /bin/caffe-master/distribute/include/caffe/common.hpp:6:0, /bin/caffe-master/distribute/include/caffe/blob.hpp:8, /bin/caffe-master/distribute/include/caffe/caffe.hpp:7, class status { /usr/local/include/tf/tensorflow/core/lib/core/status.h:37:14: error: expected unqualified-id before ‘{’ token ../me_t42.cpp:2862:1: error: expected ‘}’ @ end of input }
what seems happening both glog , tensorflow seem define same thing log(for example).
is there way solve issue? appreciated.
thank you.
so, if has issue solution this:
add
'#ifndef log_....etc'
//define state gets redefined
'#endif'
and not include
"#xlib.h"
as defines session conflicts session defined in tensorflow c++.
Comments
Post a Comment