Sabtain-Dev
/

STT-Whisper-Pashto

Automatic Speech Recognition

speech-recognition

low-resource-language

Eval Results (legacy)

Model card Files Files and versions

STT-Whisper-Pashto

Fine-tuned Whisper model for Pashto Speech Recognition.

Base Model

Whisper Small

Training Method

LoRA Fine-Tuning

Dataset

Pashto ASR Dataset

Kaggle: https://www.kaggle.com/datasets/itssabtain/pashto-asr-dataset (Recommended)
Hugging Face: https://huggingface.co/datasets/Sabtain-Dev/Pashto-ASR-Dataset

Best Result

WER: 43.83%

Downloads last month: 35

Safetensors

Model size

0.2B params

Tensor type

F32

·

Model tree for Sabtain-Dev/STT-Whisper-Pashto

Base model

openai/whisper-small

Finetuned

(3592)

this model

Dataset used to train Sabtain-Dev/STT-Whisper-Pashto

Evaluation results

Test WER on Pashto-ASR-Dataset
test set self-reported

43.830