search

lab-cosmo / metatrain

Training and evaluating machine learning models for atomistic systems.
https://lab-cosmo.github.io/metatrain/
BSD 3-Clause "New" or "Revised" License
13 stars 3 forks source link

Uniform training and validation naming #248

Closed frostedoyster closed 3 weeks ago

frostedoyster commented 3 weeks ago

Closes #240

📚 Documentation preview 📚: https://metatrain--248.org.readthedocs.build/en/248/

frostedoyster commented 3 weeks ago

@DavideTisi I think so. The idea is to use train and valid everywhere in the code, but training and validation for anything user-facing (yaml/CLI)

Luthaf commented 3 weeks ago

I don't love valid as a name, since it's not a verb/action noun like train. I could do validate if you want something shorter than validation.

frostedoyster commented 3 weeks ago

Given the context, valid is still clear. Many people even use val in the code

frostedoyster commented 3 weeks ago

https://github.com/Lightning-AI/pytorch-lightning

Luthaf commented 3 weeks ago

I think I would prefer val over valid (since it does not have a meaning on it's own), if that works for you.

frostedoyster commented 3 weeks ago

Let's go for val

frostedoyster commented 3 weeks ago

I don't think it's a good idea to change is_training to is_train because it loses some meaning (is training or is trained or ...?). Regarding the last comment, that is not user-facing from the CLI (or related yaml files).

PicoCentauri commented 3 weeks ago

Okay but then I would go for ["train", "val"]. Also, can you rebase against the main branch?