mozilla / firefox-translations-training

Training pipelines for Firefox Translations neural machine translation models
https://mozilla.github.io/firefox-translations-training/
Mozilla Public License 2.0
135 stars 28 forks source link

[meta] Make the pipeline reliable enough to train many languages #311

Open gregtatum opened 6 months ago

gregtatum commented 6 months ago

This is the lists of tasks that we need to handle in order to ramp up our ability to train many languages. These are things that break training runs, make things difficult to use the pipeline, or make it difficult for multiple people to train at once.

### Pipeline Usability
- [ ] https://github.com/mozilla/firefox-translations-training/issues/710
- [ ] https://github.com/mozilla/firefox-translations-training/issues/355
- [ ] https://github.com/mozilla/firefox-translations-training/issues/250
- [ ] https://github.com/mozilla/firefox-translations-training/issues/356
- [ ] https://github.com/mozilla/firefox-translations-training/issues/227
- [ ] https://github.com/mozilla/firefox-translations-training/issues/200
- [x] [Update worker-runner in translations generic worker image](https://mozilla-hub.atlassian.net/browse/RELOPS-782)
- [ ] https://github.com/mozilla/firefox-translations-training/issues/330
- [ ] https://github.com/mozilla/firefox-translations-training/issues/354
- [ ] https://github.com/mozilla/firefox-translations-training/issues/155
- [ ] https://github.com/mozilla/firefox-translations-training/issues/182
- [ ] https://github.com/mozilla/firefox-translations-training/issues/395
- [ ] https://github.com/mozilla/firefox-translations-training/issues/582
- [ ] https://github.com/mozilla/firefox-translations-training/issues/653
- [ ] https://github.com/mozilla/firefox-translations-training/issues/538
- [ ] https://github.com/mozilla/firefox-translations-training/issues/655
- [ ] https://github.com/mozilla/firefox-translations-training/issues/654
- [ ] https://github.com/mozilla/firefox-translations-training/issues/630
- [ ] https://github.com/mozilla/firefox-translations-training/issues/628
- [ ] https://github.com/mozilla/firefox-translations-training/issues/631
- [ ] https://github.com/mozilla/firefox-translations-training/issues/640
- [ ] https://github.com/mozilla/firefox-translations-training/issues/680
- [ ] https://github.com/mozilla/firefox-translations-training/issues/562
- [ ] https://github.com/mozilla/firefox-translations-training/issues/711
### Bugs in Training
- [ ] https://github.com/mozilla/firefox-translations-training/issues/314
- [ ] https://github.com/mozilla/firefox-translations-training/issues/293
- [ ] https://github.com/mozilla/firefox-translations-training/issues/292
- [ ] https://github.com/mozilla/firefox-translations-training/issues/272
- [ ] https://github.com/mozilla/firefox-translations-training/issues/663
- [ ] https://github.com/mozilla/firefox-translations-training/issues/649
- [ ] https://github.com/mozilla/firefox-translations-training/issues/679
bhearsum commented 3 days ago

Seeing as we're training many languages at once, should we call this done? https://github.com/mozilla/firefox-translations-training/issues/250 is the only remaining issue open in the list, and it is mostly fixed (aside from some UI jank in Taskcluster).

eu9ene commented 3 days ago

I think we're still struggling with a bunch of issues, so I'd add them here and keep this one open until we can confidently train new languages without breakage.