YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Model description

[What is this? One paragraph: fine-tuned Whisper-small for Urdu ASR, trained on Common Voice Urdu, intended to improve on base Whisper's Urdu performance.]

Intended uses & limitations

[What should someone use this for? Urdu speech transcription. What shouldn't they expect? — this was trained on read Common Voice speech; performance on spontaneous or code-switched Urdu-English speech is untested / likely worse. Say that honestly.]

Training and evaluation data

Fine-tuned on khawajaaliarshad/common-voice-urdu-processed, split by client_id (3921 train / 832 test) so no speaker appears in both splits.

Evaluation

Model WER
openai/whisper-small (base) 48.34%
Fasih779/whisper-small-urdu (this model) 34.74%

Base model evaluated on the identical held-out test split for a fair comparison.

Downloads last month
28
Safetensors
Model size
0.2B params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support