Hugging Face's logo

Spaces:
istupakov
/
onnx-asr
Running

App Files Community
onnx-asr
File size: 3,848 Bytes 
860b6bb
 
98a6067
 
 
 
 
860b6bb
e8ee9a7
 
 
 
 
 
 
 
 
 
 
 
98a6067
 
 
 
860b6bb
 
 
 
 
98a6067
 
 
 
e8ee9a7
bc2c733
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e8ee9a7
 
 
 
 
 
 
 
 
99b94cc
98a6067
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
 
from importlib.metadata import version

import gradio as gr
import numpy as np

import onnx_asr

print(f"onnx_asr version: {version('onnx_asr')}")
models = {
    name: onnx_asr.load_model(name)
    for name in [
        "gigaam-v2-ctc",
        "gigaam-v2-rnnt",
        "nemo-fastconformer-ru-ctc",
        "nemo-fastconformer-ru-rnnt",
        "alphacep/vosk-model-ru",
        "alphacep/vosk-model-small-ru",
        "whisper-base",
    ]
}


def recoginize(audio: tuple[int, np.ndarray]):
    sample_rate, waveform = audio
    try:
        waveform = waveform.astype(np.float32) / 2 ** (8 * waveform.itemsize - 1)
        return [[name, model.recognize(waveform, sample_rate=sample_rate, language="ru")] for name, model in models.items()]
    except Exception as e:
        raise gr.Error(f"{e} Audio: sample_rate: {sample_rate}, waveform.shape: {waveform.shape}.") from e


demo = gr.Interface(
    fn=recoginize,
    title="ASR demo using onnx-asr (Russian models)",
    description="""
    # ONNX ASR

    [![PyPI - Version](https://img.shields.io/pypi/v/onnx-asr.svg)](https://pypi.org/project/onnx-asr)
    [![PyPI - Downloads](https://img.shields.io/pypi/dm/onnx-asr)](https://pypi.org/project/onnx-asr)
    [![PyPI - Python Version](https://img.shields.io/pypi/pyversions/onnx-asr.svg)](https://pypi.org/project/onnx-asr)
    [![PyPI - Types](https://img.shields.io/pypi/types/onnx-asr)](https://pypi.org/project/onnx-asr)
    [![GitHub License](https://img.shields.io/github/license/istupakov/onnx-asr)](https://github.com/istupakov/onnx-asr/blob/main/LICENSE)
    [![CI](https://github.com/istupakov/onnx-asr/actions/workflows/python-package.yml/badge.svg)](https://github.com/istupakov/onnx-asr/actions/workflows/python-package.yml)

    **[onnx-asr](https://github.com/istupakov/onnx-asr)** is a Python package for Automatic Speech Recognition using ONNX models. The package is written in pure Python with minimal dependencies (*PyTorch is not required*):

    [![numpy](https://img.shields.io/badge/numpy-required-blue?logo=numpy)](https://pypi.org/project/numpy/)
    [![onnxruntime](https://img.shields.io/badge/onnxruntime-required-blue?logo=onnx)](https://pypi.org/project/onnxruntime/)
    [![huggingface-hub](https://img.shields.io/badge/huggingface--hub-optional-blue?logo=huggingface)](https://pypi.org/project/huggingface-hub/)

    ## Models used in demo:
    * `gigaam-v2-ctc` - Sber GigaAM v2 CTC ([origin](https://github.com/salute-developers/GigaAM), [onnx](https://huggingface.co/istupakov/gigaam-v2-onnx))
    * `gigaam-v2-rnnt` - Sber GigaAM v2 RNN-T ([origin](https://github.com/salute-developers/GigaAM), [onnx](https://huggingface.co/istupakov/gigaam-v2-onnx))
    * `nemo-fastconformer-ru-ctc` - Nvidia FastConformer-Hybrid Large (ru) with CTC decoder ([origin](https://huggingface.co/nvidia/stt_ru_fastconformer_hybrid_large_pc), [onnx](https://huggingface.co/istupakov/stt_ru_fastconformer_hybrid_large_pc_onnx))
    * `nemo-fastconformer-ru-rnnt` - Nvidia FastConformer-Hybrid Large (ru) with RNN-T decoder ([origin](https://huggingface.co/nvidia/stt_ru_fastconformer_hybrid_large_pc), [onnx](https://huggingface.co/istupakov/stt_ru_fastconformer_hybrid_large_pc_onnx))
    * `alphacep/vosk-model-ru` - Alpha Cephei Vosk 0.54-ru ([origin](https://huggingface.co/alphacep/vosk-model-ru))
    * `alphacep/vosk-model-small-ru` - Alpha Cephei Vosk 0.52-small-ru ([origin](https://huggingface.co/alphacep/vosk-model-small-ru))
    * `whisper-base` - OpenAI Whisper Base exported with onnxruntime ([origin](https://huggingface.co/openai/whisper-base), [onnx](https://huggingface.co/istupakov/whisper-base-onnx))
    """,
    inputs=[gr.Audio(min_length=1, max_length=20)],
    outputs=[gr.Dataframe(headers=["Model", "result"], wrap=True, show_fullscreen_button=True)],
    flagging_mode="never",
)

demo.launch()