CoolFace
Modelpublic

Quyashbek/whisper-base-karakalpak

sourceHugging Faceapache-2.0updated 1mo agoView on Hugging Face
0likes32downloads
Model Card

๐ŸŽ™๏ธ Whisper Base Karakalpak ASR

This model is a fine-tuned version of openai/whisper-base for Automatic Speech Recognition (ASR) in the Karakalpak language.


๐Ÿ‘ค Founder

Quyashbek Allanazarov


๐Ÿค Acknowledgements

  • โ€”๐ŸŽ“ New Uzbekistan University โ€” for knowledge, research environment, and academic direction
  • โ€”๐Ÿฆ Xalq Banki AI Lab โ€” for providing GPU resources and supporting the technical direction of the project

๐Ÿ“Š Model Performance

Evaluation was performed on a held-out test set.

MetricScore
WER (Word Error Rate)38.59%
CER (Character Error Rate)9.85%

Test Details

  • โ€”Total samples: 504
  • โ€”Sampling rate: 16 kHz
  • โ€”Task: Transcribe

๐Ÿงช Inference Example

python
import torch
from transformers import WhisperForConditionalGeneration, WhisperProcessor

model_id = "Quyashbek/whisper-base-karakalpak"
processor = WhisperProcessor.from_pretrained(model_id)
model = WhisperForConditionalGeneration.from_pretrained(model_id)

# speech, sr = librosa.load("audio.wav", sr=16000)
# inputs = processor(speech, sampling_rate=16000, return_tensors="pt")
# generated_ids = model.generate(inputs.input_features)
# transcription = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]