Hi,
I found the latest TensorFlow release binary is build with “–cxxopt=”-D_GLIBCXX_USE_CXX11_ABI=0" for compatibility with the older ABI, do you have plain to remove this one? that mean build with --cxxopt=“-D_GLIBCXX_USE_CXX11_ABI=1”
Hi,
I found the latest TensorFlow release binary is build with “–cxxopt=”-D_GLIBCXX_USE_CXX11_ABI=0" for compatibility with the older ABI, do you have plain to remove this one? that mean build with --cxxopt=“-D_GLIBCXX_USE_CXX11_ABI=1”
This is one of the most important issues many teams integrating TF into their projects face. Would love to hear insights from TensorFlow team on what the roadmap for this feature looks like!
A primer on that would be: why is TensorFlow still supporting GCC<5? What are the drawbacks in moving to the newer ABI?
Crossposted at:
TF v2.9 has supported the new ABI. Check Release TensorFlow 2.9.0 · tensorflow/tensorflow · GitHub