sillsdev / sillsdev/silnlp

Ramp up training for languages in the NLLB-200

Open
#280 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

needs to be revisited research
Dominant language
Python
Stars
39
Forks
7
Avg merge
1d 9h
Merged PRs (30d)
5

Description

There are some translation projects where both languages are in the NLLB-200. It may not be the best choice to take 3 drafted chapters of Mark and train the model for 20,000 steps. In that light, we should research a good path for this. Possibly this could involve investigating:

  • Train on less steps (validation split)
  • Use subset of FLORES-200 data with the new book (drag and drop into SF)
  • Use different training weights for the different text (in the SF UI?)
  • Add a tag for Scripture and a tag for "other stuff" (drop-down in the SF UI?)

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by examining the current NLLB-200 training flow and the SF UI options for validation splits, FLORES-200 subsets, training weights, and Scripture/other tags. Compare these alternatives for projects where both languages are in NLLB-200. Done means documenting a recommended path for using drafted chapters with the model.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.