Distil-Whisper Fine-tuned on TORGO Dysarthric Speech
Fine-tuned from openai/whisper-tiny on the TORGO database for improved recognition of dysarthric (motor-impaired) speech.
Training Details
- Dataset: abnerh/TORGO-database (13,239 train / 1,656 val / 1,655 test)
- Effective batch size: 16
- Learning rate: 5e-06
- Steps: 1000
- FP16 mixed precision
- Full encoder + decoder fine-tuning
Usage
from transformers import pipeline
asr = pipeline("automatic-speech-recognition", model="Menhaz/distil-whisper-torgo")
result = asr("path/to/audio.wav")
print(result["text"])
- Downloads last month
- 58
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for Menhaz/distil-whisper-torgo
Base model
openai/whisper-tiny