Automatic Speech Recognition
Transformers
TensorBoard
Safetensors
whisper
nigerian-accent
igunosa-labs
unose
Instructions to use igunosalabs/Unose-Huen-Small with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use igunosalabs/Unose-Huen-Small with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("automatic-speech-recognition", model="igunosalabs/Unose-Huen-Small")# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModelForSpeechSeq2Seq processor = AutoProcessor.from_pretrained("igunosalabs/Unose-Huen-Small") model = AutoModelForSpeechSeq2Seq.from_pretrained("igunosalabs/Unose-Huen-Small", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Unose-Huen-Small (v0.5)
Unose-Huen-Small is a fine-tuned Whisper Small model (244M parameters) optimized for Nigerian English speech recognition, low-register phonetic clarity, and colloquial cadence.
Developed by Igunosa Labs as part of the Unose speech suite.
Model Details
- Architecture: Whisper Small (Encoder-Decoder Transformer, 12 layers, 244M params)
- Validation WER: 37.07% on the Nigerian Accented English benchmark (11.31% absolute improvement over Base)
- Inference Targets: CPU / GPU / edge
- Available Formats: PyTorch safetensors, CTranslate2 (, , , )
Ecosystem Links
- Official GitHub Codebase: https://github.com/Igunosa-Labs/unose-huen
- Sister Model (Base 74M): igunosalabs/Unose-Huen-Base
- TTS Model: igunosalabs/Unose-Voice-v1.0
- Downloads last month
- -
Model tree for igunosalabs/Unose-Huen-Small
Base model
openai/whisper-small