Gigaam Rnnt

🚀 GigaAM-v2-CTC - Hugging Face transformers

This is a Russian Automatic Speech Recognition (ASR) model, GigaAM-v2-RNNT, which provides an interface with the transformers library.

🚀 Quick Start

This project offers a Russian ASR model, GigaAM-v2-RNNT, integrated with the transformers library. It allows users to perform speech recognition tasks easily.

✨ Features

• Library Integration: It has an interface with the transformers library, enabling seamless use of common transformers methods.
• Model Initialization: The model can be initialized using transformers auto classes.

📦 Installation

The following are the required library versions:

• torch 2.5.1
• torchaudio 2.5.1
• transformers 4.49.0

💻 Usage Examples

Basic Usage

from transformers import AutoModel, AutoProcessor
import torch
import torchaudio

# load audio
wav, sr = torchaudio.load("audio.wav")
# resample if necessary
wav = torchaudio.functional.resample(wav, sr, 16000)

# load model and processor
processor = AutoProcessor.from_pretrained("waveletdeboshir/gigaam-rnnt", trust_remote_code=True)
model = AutoModel.from_pretrained("waveletdeboshir/gigaam-rnnt", trust_remote_code=True)
model.eval()

input_features = processor(wav[0], sampling_rate=16000, return_tensors="pt")

# greedy prediction
with torch.no_grad():
    pred_ids = model.generate(**input_features)

# decode token ids to text
transcription = processor.batch_decode(pred_ids)[0]

📚 Documentation

This is an original GigaAM-v2-RNNT with transformers library interface. The file gigaam_transformers.py contains model, feature extractor and tokenizer classes with usual transformers methods.

📄 License

This project is licensed under the MIT license.