The problem
The links to TensorFlow installation or compilation instruction page refer to one of the latest versions of TF (which are 2.1.0 for 2.x and 1.15.0 for 1.x, at this time), while the requirements specify TensorFlow 1.10.1.
My setup
OS: Windows 10
Python: Have both 3.7 and 3.6, using a virtualenv with 3.6 inside
Issues found
Version 1.10.1 is unavailable from pip on Windows:
more readable version with black bg:
So I tried to compile TF, but with no results.
The TF build and compilation instruction page refers to a new compilation flow that uses Bazel, while versions 1.10.x required cmake, on Windows.
In addition, the pre-built packages only refer to 1.10.0, while 1.10.1 seems nowhere to be found.
Possible solutions
- Upgrade/change TF version used by this project
- Rewrite instructions in favor of a step by step working guide
- Include (in the repo) a script/requirements.txt/dockerfile/setup and/or build automation tool that automates the steps of resolving dependencies and maybe even patching TF automatically
- ... (?)
The problem
The links to TensorFlow installation or compilation instruction page refer to one of the latest versions of TF (which are 2.1.0 for 2.x and 1.15.0 for 1.x, at this time), while the requirements specify TensorFlow 1.10.1.
My setup
OS: Windows 10
Python: Have both 3.7 and 3.6, using a virtualenv with 3.6 inside
Issues found
Version 1.10.1 is unavailable from pip on Windows:
more readable version with black bg:
So I tried to compile TF, but with no results.
The TF build and compilation instruction page refers to a new compilation flow that uses Bazel, while versions 1.10.x required cmake, on Windows.
In addition, the pre-built packages only refer to 1.10.0, while 1.10.1 seems nowhere to be found.
Possible solutions