from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
import torchaudio, torch

processor = Wav2Vec2Processor.from_pretrained("abidanoaman/urdu-asr-wave2vec2-base-merged-finetuned")
model = Wav2Vec2ForCTC.from_pretrained("abidanoaman/urdu-asr-wave2vec2-base-merged-finetuned")

# Load audio (must be 16kHz mono)
waveform, sr = torchaudio.load("your_audio.wav")
inputs = processor(waveform.squeeze(), sampling_rate=16000, return_tensors="pt")

with torch.no_grad():
    logits = model(**inputs).logits

predicted_ids = torch.argmax(logits, dim=-1)
transcription = processor.decode(predicted_ids[0])
print(transcription)

Training Configuration

Epochs: 20
Batch size: 2 (grad accum: 4)
Learning rate: 0.0001
Mixed precision: True