IndicWav2Vec-Hindi

这是一个在 fairseq 中训练的 Wav2Vec2 风格的ASR模型，并移植到Hugging Face上。有关数据集、训练设置和转换为HuggingFace格式的更多详细信息，请参阅 IndicWav2Vec 存储库。注意：此模型不支持带有语言模型的推理。

运行推理的脚本

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-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

• 网站： https://ai4bharat.org/
• 代码： https://github.com/AI4Bharat
• HuggingFace： https://huggingface.co/ai4bharat