Edit model card

IndicWav2Vec-Hindi

This is a Wav2Vec2 style ASR model trained in fairseq and ported to Hugging Face. More details on datasets, training-setup and conversion to HuggingFace format can be found in the IndicWav2Vec repo.
Note: This model doesn't support inference with Language Model.

Script to Run Inference

import torch
from datasets import load_dataset
from transformers import AutoModelForCTC, AutoProcessor
import torchaudio.functional as F

DEVICE_ID = "cuda" if torch.cuda.is_available() else "cpu"
MODEL_ID = "ai4bharat/indicwav2vec-hindi"

sample = next(iter(load_dataset("common_voice", "hi", split="test", streaming=True)))
resampled_audio = F.resample(torch.tensor(sample["audio"]["array"]), 48000, 16000).numpy()

model = AutoModelForCTC.from_pretrained(MODEL_ID).to(DEVICE_ID)
processor = AutoProcessor.from_pretrained(MODEL_ID)

input_values = processor(resampled_audio, return_tensors="pt").input_values

with torch.no_grad():
    logits = model(input_values.to(DEVICE_ID)).logits.cpu()
    
prediction_ids = torch.argmax(logits, dim=-1)
output_str = processor.batch_decode(prediction_ids)[0]
print(f"Greedy Decoding: {output_str}")

About AI4Bharat

Downloads last month
1,828
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Spaces using ai4bharat/indicwav2vec-hindi 8