Indicwav2vec Odia

🚀 IndicWav2Vec - 印地語模型

本項目是一個基於 Wav2Vec2 架構的自動語音識別（ASR）模型，該模型在 fairseq 中完成訓練，並遷移至 Hugging Face。關於數據集、訓練設置以及轉換為 Hugging Face 格式的更多詳細信息，可在 IndicWav2Vec 倉庫中查看。

🚀 快速開始

運行推理的腳本

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-odia"

sample = next(iter(load_dataset("common_voice", "or", 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}")

📚 詳細文檔

模型信息

關於 AI4Bharat

• 官網：https://ai4bharat.org/
• 代碼倉庫：https://github.com/AI4Bharat
• Hugging Face 頁面：https://huggingface.co/ai4bharat