whisper-tiny-samromur-10h
Author: Páll Rúnarsson, Reykjavík University
HUMI Laboratory – Language and Voice Laboratory
Whisper-Tiny (39M) fine-tuned on a 10h nested subset of the primary Miljón/samromur-500h scaling pool.
Part of the scaling checkpoint set for "Scaling Smaller ASR Models Against Multilingual ASR Giants" (ICASSP 2027). See the paper for methodology and WER/CER results.
- Downloads last month
- 37
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support