Mike Gerber
Mike Gerber
Workaround is to install the current main branch into a temporary venv and use that to get the model: ``` # (in a temp venv) pip install git+https://github.com/qurator-spk/eynollah.git@main ocrd resmgr...
Now we had a broken `dinglehopper-line-dirs` CLI because we didn't test it...
I've put the test files I've used for manual tests here for now: https://github.com/mikegerber/dinglehopper-test-files This will need some integration in the future.
It's possible that the installation of eynollah does not conflict with sbb_binarization anymore, but with other qurator software (page2tsv etc.). I'll check.
Installing eynollah and sbb_binarization (both main/master) into one Python 3.9 venv does not cause immediately visible breakage. I can call both tools' CLIs' `--help`: | Call | result | |--|--|...
Ah, I see, this is because the revert by @bertsky is in main already. And there's not a problem in sbb_binarization anymore. So this needs more investigation. TODO: - [...
My Apache DAV service also requires digest authentication.
Thanks for this PR, it nicely illustrated some things for me. Definitely going to look into using conda myself.
This could be a dupe, because I vaguely remember asking about it before, but I may have not opened a GitHub issue yet.
Because I was complaining elsewhere: I welcome this PR, I just want to give it a proper discussion and review as it gives different CER values compared to the current...