yaroslavvb/tensorflow-community-wheels

TensorFlow 2.0.0, Python 3.7, Docker Image, no AVX / Nehalem-CPU-Instructionset | Linux x86_64

Robbi373 opened this issue · 4 comments

Robbi373 commented
  • Built from the official Docker image tensorflow/tensorflow:devel-py3 with Python3.7 instead of standard Python3.6.
  • r2.0 tensorflow release branch
  • CPU only
  • GCC version 7.4.0, but built with --cxxopt="-D_GLIBCXX_USE_CXX11_ABI=0"
  • GCC options: -march=nehalem -Wno-sign-compare (No AVX)
  • Everything other setting is standard.

Link to wheel: https://drive.google.com/open?id=1fTAbxu99olgmb8WrxurO8ovOGR0VwzED

pavelzamyatin commented

Unfortunately couldn't get it working on my Ubuntu 16.04 due to ImportError: /lib/x86_64-linux-gnu/libm.so.6: version 'GLIBC_2.27' not found. Thanks anyway.

Robbi373 commented

Seems like the official docker image is a newer Linux version. I'm not even sure Python 3.7 runs on Ubuntu 16.04. If you can work with Python 3.5, have a look at this issue: #123 .

DSLituiev commented

keep getting ModuleNotFoundError: No module named 'tensorflow.python.tools'; 'tensorflow.python' is not a package on Ubuntu 18

Robbi373 commented

keep getting ModuleNotFoundError: No module named 'tensorflow.python.tools'; 'tensorflow.python' is not a package on Ubuntu 18

This does not seem to be an issue with my wheel, but rather with your tensorflow installation(s) and/or file naming. Just look for that error with your favourite search engine.