google/fleurs
Viewer • Updated • 768k • 103k • 450
openai/whisper-small fine-tuned for Hausa automatic speech recognition on the
ha_ng config of google/fleurs
(3259 train / 621 test clips), 3 epochs.
Test set WER: 44.68% (eval_loss: 0.712)
from transformers import WhisperForConditionalGeneration, WhisperProcessor
processor = WhisperProcessor.from_pretrained("nahomazmach/whisper-small-ha")
model = WhisperForConditionalGeneration.from_pretrained("nahomazmach/whisper-small-ha")
Base model
openai/whisper-small