Fallovski/french-serer-nllb-full-ft

French–Serer NLLB Full Fine-Tuning

Part of a benchmark of six configurations for French→Serer neural machine translation. Serer is a critically low-resource Niger-Congo language (~1.2M speakers, Senegal/Gambia), phylogenetically close to the well-resourced Wolof.

Model summary

• —Experiment ID: B
• —Kind: finaltranslationmodel
• —Direction: French → Serer
• —Base model: facebook/nllb-200-distilled-600M (revision: main)
• —Best checkpoint: french_serer_nllb_full_ft-epoch=08-val_bleu=22.4582.ckpt
• —Random seed: 42

Hyperparameters

• —LR: 2e-05
• —NUM_EPOCHS: 10
• —BATCH_SIZE: 8
• —GRAD_ACCUM_STEPS: 4
• —WARMUP_STEPS: 500

Critical limitation — Wolof decoding proxy

Serer (srr_Latn) is not a supported NLLB-200 target. This model decodes under the wol_Latn (Wolof) language tag as a proxy, fine-tuned on French–Serer data. Automatic metrics (especially BLEU) can be partly inflated by lexical/orthographic overlap with Wolof; see the companion proximity-probe calibration for this corpus (Fallovski/french-serer-nllb-wolof-proximity-probe). Human review by a qualified Serer speaker is strongly recommended before any downstream use.

Intended use

Research on French-to-Serer machine translation on a corpus that is ~90% religious (Bible) register, ~10% educational glossaries, primarily Siin dialect. Not validated for legal, medical, emergency, or fully autonomous publication use. Private repository — not intended for public deployment in its current state.

Evaluation

Evaluated on the held-out test split (2890 sentence pairs, SHA-256 of the split: 01d14d982a3c0cce172b5099e2db064d05bdef89677fe72a45f56715d6364ee2). Metrics were computed with the project's own evaluation scripts (not copied from the manuscript without independent reproduction); the training and evaluation code is kept in a private repository, available on request.

Training data and rights

Parallel corpus of 23113 train / 2889 val / 2890 test French–Serer sentence pairs, built primarily from religious texts (Bible, ~90%) and educational glossaries (~10%), predominantly Siin dialect. Preprocessing: Unicode normalization, exact-duplicate removal, length-ratio filtering (1:3–3:1). Document-level splitting was not possible (no document identifiers available); the split is at the sentence level with a fixed seed. Full provenance, licensing, and consent documentation are kept in a private dataset card, available on request, prior to any public release.