Instructions to use malinali-app/traduction-en-wolof with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use malinali-app/traduction-en-wolof with Transformers:
# Use a pipeline as a high-level helper # Warning: Pipeline type "translation" is no longer supported in transformers v5. # You must load the model directly (see below) or downgrade to v4.x with: # pip install "transformers<5.0.0" from transformers import pipeline pipe = pipeline("translation", model="malinali-app/traduction-en-wolof")# pip install -U transformers accelerate # Load model directly from transformers import AutoTokenizer, AutoModelForSeq2SeqLM tokenizer = AutoTokenizer.from_pretrained("malinali-app/traduction-en-wolof") model = AutoModelForSeq2SeqLM.from_pretrained("malinali-app/traduction-en-wolof", device_map="auto") - Notebooks
- Google Colab
- Kaggle
English → Wolof Marian / Candle (Malinali)
Public on-device pack for Malinali. Fine-tuned MarianMT
weights in safetensors form plus fast tokenizers for Candle (marian_flutter).
Upstream credit
- Fine-tune:
LocaleNLP/eng_wolof(MIT) - Base model:
Helsinki-NLP/opus-mt-en-mul - Base project: Helsinki-NLP / OPUS-MT
Malinali only repackages the root weights and converts SentencePiece to Hugging Face fast tokenizer JSON. We do not claim ownership of the trained model.
This pack is a dedicated English→Wolof fine-tune of opus-mt-en-mul, not a
general multilingual model. Always prefix >>wol<< on the source text
(token id 1136). Malinali sets sourcePrefix in the catalogue.
Tokenizer
source.spm, target.spm, and vocab.json are the opus-mt-en-mul pair
(vocab size 64110, pad id 64109). >>wol<< is id
1136. tokenizer_config.json still says target_lang: mul;
do not trust that label — pin Wolof via the prefix.
Files (required)
| File | Role |
|---|---|
config.json |
Marian config |
model.safetensors |
Weights (~308 Mo F32) |
tokenizer-enc.json |
English (source) fast tokenizer |
tokenizer-dec.json |
Target fast tokenizer (shared mul vocab) |
Direction: English → Wolof.
Quality
Upstream self-reported BLEU 76.12 on their custom ~84k English–Wolof dataset. Malinali shows this as BLEU 76.1 / 100. This is not an OPUS / SacreBLEU score computed by Malinali.
- Downloads last month
- 33
Model tree for malinali-app/traduction-en-wolof
Base model
Helsinki-NLP/opus-mt-en-mulEvaluation results
- BLEU on English-Wolof Custom Datasetself-reported76.120