search

OpenBioLink / ThoughtSource

A central, open resource for data and tools related to chain-of-thought reasoning in large language models. Developed @ Samwald research group: https://samwald.info/
MIT License
887 stars 72 forks source link

Restrict python>=3.7,<3.11 in conda.yml to be consistent with setup.cfg #89

Closed jas-ho closed 1 year ago

jas-ho commented 1 year ago

Without this we get the following error when attempting to create a conda env from conda.yml: ERROR: Package 'cot' requires a different Python: 3.11.0 not in '<3.11,>=3.7'

KonstantinHebenstreit commented 1 year ago

Thanks jas-no! Your suggestion is of course correct. The conda.yml file is a leftover, we decided for pip venv so far. Installation procedure is described in the readme. I will delete the conda yml file for now, that others do not run in the same error. I will open an issue for making it also run in conda, but I think that needs some more changes than only the python version unfortunately. If you want to do that together, I am up for it, but there is some other improvements I want to do before that.

matthias-samwald commented 1 year ago

BTW running "pip install -e ./libs/cot[api]" in an existing conda environment worked for me last time I tried, so I guess it "works with conda" if that means running pip in a conda env

KonstantinHebenstreit commented 1 year ago

Yes, that is true. Just having a conda.yml file in our repo to build a conda environment was misleading...

KonstantinHebenstreit commented 1 year ago

We can also leave the conda.yml file there. I just never did that, to have a conda.yml file which just calls a pip install of all other packages. If that makes sense, we can totally do that! @jas-ho What is your opinion on that?

jas-ho commented 1 year ago

no, removing conda.yml seems like the best choice. anyone is still free to do what @matthias-samwald mentioned (setting up a suitable conda env and then running the pip install from within).

jas-ho commented 1 year ago

(just for context: I take the existence of a conda.yml file as a hint that I will get a fully-working environment in a single call. but that is not necessary of course and probably not worth maintaining a partially redundant double structure between conda and setup_tools).