0

I am aware that Tensorflow v1.5+ comes shipped with Eager execution, however, I can't install any version higher than v1.4 due to its CUDA/CuDNN requirements (https://www.tensorflow.org/install/source#tested_build_configurations). Specifically, version CUDA v9 does not support Ubuntu 14.04 and I am unable to simply upgrade my operating system.

I have seen that you could use the nightly builds (Error in tensorflow eager module) when v1.4 was most recent. Is there any way to conda/pip install an older build of 1.4 with eager execution enabled? If I simply pip install the gpu nightly build it will install the most recent Tensorflow version.

talonmies
  • 70,661
  • 34
  • 192
  • 269
Roy
  • 3,027
  • 4
  • 29
  • 43
  • The latest stable tensorflow version is 1.12 – Amir Jan 03 '19 at 11:55
  • 1
    Right sure, but I can't use that version because of its dependence on CUDA v9. I'm wondering if there is a way of using eager execution in v1.4. People have done it using the experimental nightly builds but I'm not sure how. – Roy Jan 03 '19 at 12:07
  • 1
    Before blindly requesting that this question should be closed at least provide a constructive comment so I can improve the question.. – Roy Jan 03 '19 at 12:11
  • With regards to the close request. I am not asking for recommendations on any external resources/books etc Can you please read the question first. I have covered the issue I am facing and why, and what has been covered elsewhere of which is no longer relevant. – Roy Jan 03 '19 at 13:24

0 Answers0