Instructions to use goaicorp/dyu-whisper-base with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use goaicorp/dyu-whisper-base with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("automatic-speech-recognition", model="goaicorp/dyu-whisper-base")# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModelForSpeechSeq2Seq processor = AutoProcessor.from_pretrained("goaicorp/dyu-whisper-base") model = AutoModelForSpeechSeq2Seq.from_pretrained("goaicorp/dyu-whisper-base", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Request access to dyu-whisper-base
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
dyu-whisper-base is shared under CC BY-SA 4.0: it was trained on Koumankan4Dyula by UVCI and Data354. Please read the card, its terms and licenses before requesting access, and make sure you understand them.
Log in or Sign Up to review the conditions and access this model content.
dyu-whisper-base
Dioula speech recognition: Whisper base fine-tuned by GO AI Corp.
| Task | Speech recognition, Dioula (dyu) |
| Base model | openai/whisper-base |
| Training data | dyu-speech train (Koumankan; inferred from the number of training steps) |
| Published | April 2026 |
Résumé en français. Reconnaissance de la parole en dioula : Whisper base affiné par GO AI Corp. Entraîné sur l'ensemble
traindedyu-speech(Koumankan ; déduit du nombre d'étapes d'entraînement). Publié en avril 2026. Licence CC BY-SA 4.0, comme ses données d'entraînement : citez UVCI et Data354 (Koumankan4Dyula) et partagez vos adaptations sous la même licence. Accès sur demande.
Quick start
from transformers import pipeline
asr = pipeline("automatic-speech-recognition", model="goaicorp/dyu-whisper-base")
print(asr("audio.wav")["text"]) # 16 kHz mono works best
Evaluation
| Test set | WER | CER |
|---|---|---|
| dyu-speech test (1,393 clips) | pending | pending |
The run reported WER 106.6% on its own validation split, which is not comparable with other systems.
Limits
- Early model: word error rates above 80% make it unusable in practice.
- Read speech only.
License
CC BY-SA 4.0, because it was trained on CC BY-SA data (Koumankan4Dyula, UVCI and Data354). Credit them and share adaptations under the same license.
Citation
@misc{goaicorp2026dyuwhisperbase,
title = {dyu-whisper-base: Dioula speech recognition with Whisper base},
author = {Nyantudre, Wendyellé Abubakrh Alban and Bandaogo, Aristide},
year = {2026},
howpublished = {Hugging Face Models, GO AI Corp},
url = {https://huggingface.co/goaicorp/dyu-whisper-base}
}
Developed and maintained by
- Wendyellé A. Alban NYANTUDRE: nyantudre@gmail.com
- Aristide BANDAOGO: aristide@goaicorporation.org
- Downloads last month
- 8
Model tree for goaicorp/dyu-whisper-base
Base model
openai/whisper-base