You are not logged in.
Pages: 1
Hello,
The tensorflow package (specifically: community/tensorflow-opt-cuda which I am using) is missing headers.
A very sample example cannot be compiled, and I doubt the package is any useful in its current form.
Some headers are there, but many are not!
For example:
the header /usr/include/tensorflow/core/lib/core/status.h includes absl/strings/string_view.h which does not exist (can be solved by installing abseil-cpp from aur)
the header /usr/include/tensorflow/core/framework/tensor.h includes third_party/eigen3/unsupported/Eigen/CXX11/Tensor which does not exist
Of course, ultimately, the dev can simply clone the tensorflow git and add it to his include path. But, in such case I don't see the reason why the tensorflow package includes so many header files when installed !?
Sample code:
#include <tensorflow/core/platform/env.h>
#include <tensorflow/core/public/session.h>
#include <iostream>
using namespace std;
using namespace tensorflow;
int main() {
Session* session;
Status status = NewSession(SessionOptions(), &session);
if (!status.ok()) {
cout << status.ToString() << "\n";
return 1;
}
cout << "Session successfully created.\n";
}Notice: this is not a coding / programming issue. The code sample has just been provided to illustrate the issue with the tensorflow package.
Edit: All the necessary headers are available in the package python-tensorflow (in /usr/lib/python3.7/site-packages/tensorflow/include/). The tensorflow package is confusing me even more.
Last edited by samarch (2019-02-27 00:30:41)
Offline
You should rather file a bug against the package(s), judging from https://bugs.archlinux.org/index.php?st … =&do=index this seems a common pattern w/ this split package.
Offline
Pages: 1