Downloads · 30 days
37
100% of all-time downloads
quechuabert
makitty/quechuabert
quechuabert is a fill-mask model from makitty. Use it when you need the model to fill a missing word. It is set up for transformers. The card lists the license as mit.
Small BertForMaskedLM for Southern Quechua, tokenized with the morphological PRPE segmenter from QuechuaTok.
All-time downloads
37
Public
Parameters
4.2M
17 MB on disk
Likes
1
Public
Repo makeup
Click a slice to open those files.
.safetensors17 MB · 99%
At a glance
- Task
- Fill-Mask
- Library
- transformers
- License
- mit
- Model type
- bert
- Access
- Public
- Created
- Sep 7, 2026
- Updated
- Sep 7, 2026
- SHA
- 15c6a3c0
Datasets
Model card
From the Hugging Face model README
QuechuaBERT (PRPE)
Small BertForMaskedLM for Southern Quechua, tokenized with the morphological PRPE segmenter from QuechuaTok.
Training note
CPU run: ~20k Llamacha lines, 5000 MLM steps, tiny BERT (hidden=256, layers=4). Final train loss ≈ 6.19. Early checkpoint for visibility, not a finished SOTA model.
Paper
Contreras, M. (2026). QuechuaTok. arXiv:2606.23943 — https://arxiv.org/abs/2606.23943
Usage
Install QuechuaTok, then load the model weights from this repo. Custom PRPE tokenizer lives in quechuatok.hf_tokenizer.
See QuechuaTok issue #8 and scripts/train_quechuabert.py.
- Task
- Fill-Mask
- Library
- transformers
- Type
- bert
- License
- mit
- Languages
- qu
- Created
- Sep 7, 2026
- Updated
- Sep 7, 2026