main.py

#!/usr/bin/python3
# -*- coding: utf-8 -*-
import argparse

import gradio as gr

from examples import examples
from models import model_map
from project_settings import project_path


def get_args():
    parser = argparse.ArgumentParser()
    parser.add_argument(
        "--examples_dir",
        default=(project_path / "data/examples").as_posix(),
        type=str
    )
    parser.add_argument(
        "--trained_model_dir",
        default=(project_path / "trained_models").as_posix(),
        type=str
    )
    args = parser.parse_args()
    return args


def update_model_dropdown(language: str):
    if language not in model_map.keys():
        raise ValueError(f"Unsupported language: {language}")

    choices = model_map[language]
    choices = [c["repo_id"] for c in choices]
    return gr.Dropdown(
        choices=choices,
        value=choices[0],
        interactive=True,
    )


def build_html_output(s: str, style: str = "result_item_success"):
    return f"""
    <div class='result'>
        <div class='result_item {style}'>
          {s}
        </div>
    </div>
    """


def process_uploaded_file(language: str,
                          repo_id: str,
                          decoding_method: str,
                          num_active_paths: int,
                          add_punctuation: str,
                          in_filename: str,
                          ):
    return "Dummy", build_html_output("Dummy")


def main():
    title = "# Automatic Speech Recognition with Next-gen Kaldi"

    language_choices = ["Chinese"]

    language_to_models = {
        "Chinese": ["None"]
    }

    # components
    language_radio = gr.Radio(
        label="Language",
        choices=language_choices,
        value=language_choices[0],
    )
    model_dropdown = gr.Dropdown(
        choices=language_to_models[language_choices[0]],
        label="Select a model",
        value=language_to_models[language_choices[0]][0],
    )
    decoding_method_radio = gr.Radio(
        label="Decoding method",
        choices=["greedy_search", "modified_beam_search"],
        value="greedy_search",
    )
    num_active_paths_slider = gr.Slider(
        minimum=1,
        value=4,
        step=1,
        label="Number of active paths for modified_beam_search",
    )
    punct_radio = gr.Radio(
        label="Whether to add punctuation (Only for Chinese and English)",
        choices=["Yes", "No"],
        value="Yes",
    )

    # blocks
    with gr.Blocks() as blocks:
        gr.Markdown(value=title)

        with gr.Tabs():
            with gr.TabItem("Upload from disk"):
                uploaded_file = gr.Audio(
                    sources=["upload"],
                    type="filepath",
                    label="Upload from disk",
                )
                upload_button = gr.Button("Submit for recognition")
                uploaded_output = gr.Textbox(label="Recognized speech from uploaded file")
                uploaded_html_info = gr.HTML(label="Info")

                gr.Examples(
                    examples=examples,
                    inputs=[
                        language_radio,
                        model_dropdown,
                        decoding_method_radio,
                        num_active_paths_slider,
                        punct_radio,
                        uploaded_file,
                    ],
                    outputs=[uploaded_output, uploaded_html_info],
                    fn=process_uploaded_file,
                )

        language_radio.change(
            update_model_dropdown,
            inputs=language_radio,
            outputs=model_dropdown,
        )

    blocks.queue().launch()

    return


if __name__ == "__main__":
    main()