Paul Henderson

I think this is not specific to DIRT, but a problem with your cuda compiler -- it looks like your version of nvcc is not compatible with your version of...

@MISSXUAN0 What versions of gcc, nvcc, and cmake are you using? I cannot reproduce this. See https://github.com/highfidelity/hifi/issues/8047 and https://github.com/mapillary/inplace_abn/issues/106 for a couple of possible causes & fixes.

@MISSXUAN0 The output shows your problem is with tensorflow's `string_view.h`, not the same as the problem with `math.h` in this thread. See https://github.com/pmh47/dirt/issues/6, and https://github.com/tensorflow/tensorflow/issues/22766 for workarounds.

@MISSXUAN0 See https://github.com/pmh47/dirt/issues/10, the last two comments there give the cause and fix for that problem

That should have worked, assuming you installed in development mode afterwards (i.e. `pip install -e .`). Try deleting the build folder, making sure DIRT has been fully uninstalled from your...

@yyvettey Add `-D_GLIBCXX_USE_CXX11_ABI=0` to `CMAKE_CXX_FLAGS` in build/CMakeCache.txt -- it's a flag to g++, not to cmake. Alternatively, export `CXXFLAGS='-D_GLIBCXX_USE_CXX11_ABI=0'` before running cmake for the first time.

@wkailiu For EGL libraries not found, see #17

@haohao-hu Unfortunately DIRT does not support Windows.

Thanks for the contribution @Timen -- I'll test it with some other versions of tensorflow before merging, to check everything is still ok there. In the changes to the samples,...

Sorry for the very slow reply! Yes, DIRT itself works in tf 2.0/2.1; 91a05fc fixes `tests/square_test.py` to be tf2 compatible, meaning the standard install-and-test procedure now works for 1.x or...