can we modify “pip installed torch” to use _GLIBCXX_USE_CXX11_ABI=1 for torch compilation ? It seems all pip installed torch comes with _GLIBCXX_USE_CXX11_ABI=0The reason is torch compiled model using AOT (Ahead of time compilation) doesn’t work in c++ inference code which is compiled with ABI=1.
All of our binaries should support CXX11 ABI for a few releases already after CentOS7 was dropped as the build env. Which release are you using?
Thank you for the reply. I was using torch==2.6.0+cu124
After using torch==2.9.1, I don’t see any issue anymore.