Sentence Similarity
sentence-transformers
ONNX
Safetensors
Morisyen
English
French
xlm-roberta
feature-extraction
mauritian-creole
kreol-morisien
matryoshka
mteb
Eval Results (legacy)
text-embeddings-inference
Instructions to use Singaraj/morisien-embed with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- sentence-transformers
How to use Singaraj/morisien-embed with sentence-transformers:
from sentence_transformers import SentenceTransformer model = SentenceTransformer("Singaraj/morisien-embed") sentences = [ "The weather is lovely today.", "It's so sunny outside!", "He drove to the stadium." ] embeddings = model.encode(sentences) similarities = model.similarity(embeddings, embeddings) print(similarities.shape) # [3, 3] - Notebooks
- Google Colab
- Kaggle
Commit History
19 days ago
-
Cite MorisienMT for the register line
Verified signaturedaa39be
21 days ago
-
Drop unscripted figures and the withdrawn bound
Verified signature25d71b3
-
Point to v1.5 and correct three claims the preprint revision withdrew
Verified signatureWeights are untouched. This is the checkpoint in the preprint and in the MTEB task, and it stays available unchanged. Adds a pointer to morisien-embed-v1.5, which is fine-tuned from LaBSE and scores better out of domain, where this model does not beat untrained LaBSE. Corrects three statements that the revised preprint (10.5281/zenodo.22640806) withdraws. The corpus was described as effectively all publicly available Creole parallel text; google/smol holds 2,472 further gold pairs it does not contain. The leak filters were described as removing evaluation data; they drop 0 of 69,525 raw rows because the splits were already disjoint. And the FLORES+ ceiling was read as evidence of generalisation; a TF-IDF baseline scores 0.8221 on that pool, so it measures surface overlap.d50c668
-
Update README.md
Verified signatureAdd MTEB results and paper DOI to model card93f9a6d
-
MorisienMT relicensed to MIT
Verified signature3187caa -
Add int8 ONNX weights for in-browser inference (transformers.js)
Verified signature875b3c6 -
Add model-index eval results to card metadata
Verified signaturea6a1252 -
Add English->Creole results; release wording
Verified signature9ca09f9 -
Haitian figures from the committed probe; STS-b and truncation limitations
Verified signaturefe1f289
-
Canonical dataset id in metadata
Verified signature55c2406 -
Protocol note precision
Verified signature9c7c1be -
Precision fixes from claims audit
Verified signature4d69e01 -
Sync model card: reproducibility scoping, seed-protocol precision
Verified signature75ad30f -
Replace auto-generated card with the full model card
Verified signaturebd84088 -
Add new SentenceTransformer model
Verified signature6c1469f -
initial commit
Verified signature54d11d5