JonaWhisper/kenlm-models
Updated
AI & ML interests
None defined yet.
Recent Activity
Organization Card
JonaWhisper
Converted models for JonaWhisper, an open-source local-first dictation app.
Models
Grammar & Spelling Correction (ONNX)
Converted from PyTorch sources, quantized to INT8 for fast on-device inference.
| Model | Task | Params | Languages | Formats |
|---|---|---|---|---|
| jonawhisper-gec-t5-small-onnx | Grammar correction | 60M | 11 (multilingual) | FP32 + INT8 |
| jonawhisper-t5-spell-fr-onnx | Spell correction | 220M | French | FP32 + INT8 |
| jonawhisper-flanec-base-onnx | Grammar correction | 250M | English | FP32 + INT8 |
| jonawhisper-flanec-large-onnx | Grammar correction | 800M | English | FP32 + INT8 |
Language Models (KenLM)
| Model | Description | Languages |
|---|---|---|
| kenlm-models | Pruned trigram models trained on Wikipedia | FR, EN, DE, ES, PT, IT, NL, PL, RU |
datasets 0
None public yet