tabs/cls_tab.py

#!/usr/bin/python3
# -*- coding: utf-8 -*-
import argparse
import json
from functools import lru_cache, partial
from pathlib import Path
import shutil
import tempfile
import time
import zipfile
from typing import Tuple

import gradio as gr
import torch

from project_settings import project_path
from toolbox.torch.utils.data.vocabulary import Vocabulary


@lru_cache(maxsize=100)
def load_model(model_file: Path):
    with zipfile.ZipFile(model_file, "r") as f_zip:
        out_root = Path(tempfile.gettempdir()) / "cc_audio_8"
        if out_root.exists():
            shutil.rmtree(out_root.as_posix())
        out_root.mkdir(parents=True, exist_ok=True)
        f_zip.extractall(path=out_root)

    tgt_path = out_root / model_file.stem
    jit_model_file = tgt_path / "trace_model.zip"
    vocab_path = tgt_path / "vocabulary"

    vocabulary = Vocabulary.from_files(vocab_path.as_posix())

    with open(jit_model_file.as_posix(), "rb") as f:
        model = torch.jit.load(f)
    model.eval()

    shutil.rmtree(tgt_path)

    d = {
        "model": model,
        "vocabulary": vocabulary
    }
    return d


def when_click_cls_button(audio_t,
                          model_name: str,
                          ground_true: str) -> Tuple[str, float]:

    sample_rate, signal = audio_t

    model_file = project_path / f"trained_models/{model_name}.zip"
    d = load_model(model_file)

    model = d["model"]
    vocabulary = d["vocabulary"]

    inputs = signal / (1 << 15)
    inputs = torch.tensor(inputs, dtype=torch.float32)
    inputs = torch.unsqueeze(inputs, dim=0)

    time_begin = time.time()
    with torch.no_grad():
        logits = model.forward(inputs)
        probs = torch.nn.functional.softmax(logits, dim=-1)
        label_idx = torch.argmax(probs, dim=-1)
    time_cost = time.time() - time_begin

    label_idx = label_idx.cpu()
    probs = probs.cpu()

    label_idx = label_idx.numpy()[0]
    prob = probs.numpy()[0][label_idx]

    label_str = vocabulary.get_token_from_index(label_idx, namespace="labels")

    result = {
        "label": label_str,
        "prob": round(float(prob), 4),
        "time_cost": round(time_cost, 4),
    }
    result = json.dumps(result, ensure_ascii=False, indent=4)
    return result


def when_model_name_change(model_name: str, cls_trained_model_dir: Path):
    m = load_model(
        model_file=(cls_trained_model_dir / f"{model_name}.zip")
    )
    token_to_index: dict = m["vocabulary"].get_token_to_index_vocabulary(namespace="labels")
    label_choices = list(token_to_index.keys())

    split_label = gr.Dropdown(choices=label_choices, value=label_choices[0], label="label")

    return split_label


def get_cls_tab(examples_dir: str, trained_model_dir: str):
    cls_examples_dir = Path(examples_dir)
    cls_trained_model_dir = Path(trained_model_dir)

    # models
    cls_model_choices = list()
    for filename in cls_trained_model_dir.glob("*.zip"):
        model_name = filename.stem
        if model_name == "examples":
            continue
        cls_model_choices.append(model_name)
    model_choices = list(sorted(cls_model_choices))

    # model_labels_choices
    m = load_model(
        model_file=(cls_trained_model_dir / f"{model_choices[0]}.zip")
    )
    token_to_index = m["vocabulary"].get_token_to_index_vocabulary(namespace="labels")
    model_labels_choices = list(token_to_index.keys())

    # examples zip
    cls_example_zip_file = cls_trained_model_dir / "examples.zip"
    with zipfile.ZipFile(cls_example_zip_file.as_posix(), "r") as f_zip:
        out_root = cls_examples_dir
        if out_root.exists():
            shutil.rmtree(out_root.as_posix())
        out_root.mkdir(parents=True, exist_ok=True)
        f_zip.extractall(path=out_root)

    # examples
    cls_examples = list()
    for filename in cls_examples_dir.glob("**/*/*.wav"):
        label = filename.parts[-2]
        cls_examples.append([
            filename.as_posix(),
            model_choices[0],
            label
        ])

    with gr.TabItem("cls"):
        with gr.Row():
            with gr.Column(scale=3):
                cls_audio = gr.Audio(label="audio")
                with gr.Row():
                    cls_model_name = gr.Dropdown(choices=model_choices, value=model_choices[0], label="model_name")
                    cls_label = gr.Dropdown(choices=model_labels_choices, value=model_labels_choices[0], label="label")
                with gr.Row():
                    cls_ground_true = gr.Textbox(label="ground_true")
                    cls_button = gr.Button("run", variant="primary")

            with gr.Column(scale=3):
                cls_outputs = gr.Textbox(label="outputs")

        gr.Examples(
            cls_examples,
            inputs=[cls_audio, cls_model_name, cls_ground_true],
            outputs=[cls_outputs],
            fn=when_click_cls_button,
            examples_per_page=5,
        )

        cls_model_name.change(
            partial(when_model_name_change, cls_trained_model_dir=cls_trained_model_dir),
            inputs=[cls_model_name],
            outputs=[cls_label],
        )

        cls_button.click(
            when_click_cls_button,
            inputs=[cls_audio, cls_model_name, cls_ground_true],
            outputs=[cls_outputs],
        )

    return locals()


if __name__ == "__main__":
    pass