axondeepseg: No matching distribution found for tensorflow==1.3.0

using 2bee818b5be963b11f57733b110f1818daebf402 on rosenberg, i cannot properly install tensorflow==1.3.0:

[...]
Collecting tensorflow==1.3.0 (from AxonDeepSeg==2.2.dev0)
  Could not find a version that satisfies the requirement tensorflow==1.3.0 (from AxonDeepSeg==2.2.dev0) (from versions: )
No matching distribution found for tensorflow==1.3.0 (from AxonDeepSeg==2.2.dev0)
(venv_ads) [jcohen@rosenberg axondeepseg]$ pip install tensorflow==1.3.0
Collecting tensorflow==1.3.0
  Could not find a version that satisfies the requirement tensorflow==1.3.0 (from versions: )
No matching distribution found for tensorflow==1.3.0
(venv_ads) [jcohen@rosenberg axondeepseg]$ pip -V
pip 18.1 from /home/jcohen/miniconda3/envs/venv_ads/lib/python3.7/site-packages/pip (python 3.7)
(venv_ads) [jcohen@rosenberg axondeepseg]$ python
Python 3.7.1 (default, Oct 23 2018, 19:19:42) 
[GCC 7.3.0] :: Anaconda, Inc. on linux

About this issue

  • Original URL
  • State: closed
  • Created 6 years ago
  • Comments: 30 (28 by maintainers)

Most upvoted comments

yup! i confirm: it’s a 3.7 issue. Works with 3.6

I encountered the same issue while using anaconda with python 3.7.1. I was able to resolve it by creating a new conda virtual env with python 3.6

i would like to explore possibilities with miniconda, because it is much lighter (not for me, but for potential ADS users). Maybe the issue is related to 3.7 (https://github.com/tensorflow/tensorflow/issues/17022#issuecomment-435357209). Exploring…

Thanks for the note @Evans-Okerosi ! We do in fact specify in the documentation (here) to use a virtual environment with Python 3.6. Did you install using pip? That’s always been a worry of mine; users installing using that command are likely to miss the documentation that we post on GitHub.

As @mathieuboudreau mentioned, I would also lean towards not supporting older OS, simply because we don’t have the bandwidth. Also I would be more aggressive than just showing a “warning” (that only 10% people will see, and those who will see it will forget its existence 5min later).

Instead, I would check for the platform during installation, and if it does not fit OS specs, I would exit install with error, saying that “OS is not supported, please install more recent version, see doc: URL_DOC”.

Yeah, I haven’t been able to create a good virtualenv with conda yet on Rosenberg. Maybe @perone can chime in again.

The virtual environment steps that ended up working for me last week (passing pip installation & passing unit tests) were:

  • Go to the user’s home directory: cd ~
  • python3.6 -m virtualenv venv_ads
  • source venv_ads/bin/activate