Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

update submodules (avoiding TF 1 vs 2 conflict) #76

Merged
merged 1 commit into from
Apr 8, 2020

Conversation

bertsky
Copy link
Collaborator

@bertsky bertsky commented Apr 8, 2020

This brings updates to ocrd_keraslm and cor-asv-ann which will completely avoid conflict with TF2 ocrd_anybaseocr – as a temporary solution to #69.

@bertsky bertsky requested a review from kba April 8, 2020 13:03
@bertsky bertsky force-pushed the update-to-avoid-tf2-calamity branch from cfecf45 to b404c9e Compare April 8, 2020 13:11
@stweil stweil merged commit bc9c9b3 into OCR-D:master Apr 8, 2020
@bertsky bertsky deleted the update-to-avoid-tf2-calamity branch April 8, 2020 13:49
@stweil
Copy link
Collaborator

stweil commented Apr 8, 2020

@bertsky, all recent Linux distributions with Python 3.8 might fail with make all:

ERROR: No matching distribution found for tensorflow-gpu==1.15.* (from ocrd-cor-asv-ann==0.1.2)

I use Debian Bullseye (=testing) with Python 3.8.2.

@bertsky
Copy link
Collaborator Author

bertsky commented Apr 8, 2020

all recent Linux distributions with Python 3.8 might fail with make all:

ERROR: No matching distribution found for tensorflow-gpu==1.15.* (from ocrd-cor-asv-ann==0.1.2)

I use Debian Bullseye (=testing) with Python 3.8.2.

Gosh, these TF people drive me crazy. First they just redefine their naming scheme twice for the same versions (tensorflow vs tensorflow-gpu on pre-2), then they don't even provide up-to-date builds anymore.

But you have to concede, they say so directly on their website:

TensorFlow is tested and supported on the following 64-bit systems:

Python 3.5–3.7

And the way they say it, it sounds they will provide 3.8 support...

@stweil
Copy link
Collaborator

stweil commented Apr 8, 2020

We have several ways how to handle this:

  • Do nothing. It will be solved as soon as there is a TF binary for Python 3.8.
  • Do nothing. It only affects people who use the latest distributions. Those people should know how to handle problems.
  • Add documentation to the README and hope that people read it.
  • Enhance the Makefile to check the Python version. Fail with an error message if it is 3.8, or look whether there is also a Python 3.6 or 3.7 available and use that.
  • ...

@kba
Copy link
Member

kba commented Apr 8, 2020

I'm for a combination of "Do nothing" and a warning for python 3.8 users in the README. It's imperative for TF to provide up-to-date versions, supporting 3.8 must be high on their priorities and users that are on bleeding edge distros like Debian 11 or Ubuntu 20.04 can be expected to check the documentation or even install an older python in case of problems.

@bertsky
Copy link
Collaborator Author

bertsky commented Apr 8, 2020

On the other hand, we don't even look for Python 3.8 support in core ourselves yet (CircleCI and Travis use 3.5-3.7). That's not "3.5 or higher"...

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants