app.py

import re
import os
import time
import torch
import shutil
import argparse
import warnings
import gradio as gr
from transformers import GPT2Config
from model import Patchilizer, TunesFormer
from convert import abc2xml, xml2, xml2img
from utils import (
    PATCH_NUM_LAYERS,
    PATCH_LENGTH,
    CHAR_NUM_LAYERS,
    PATCH_SIZE,
    SHARE_WEIGHTS,
    WEIGHTS_PATH,
    TEMP_DIR,
    TEYVAT,
    DEVICE,
    EN_US,
    _L,
)


def get_args(parser: argparse.ArgumentParser):
    parser.add_argument(
        "-num_tunes",
        type=int,
        default=1,
        help="the number of independently computed returned tunes",
    )
    parser.add_argument(
        "-max_patch",
        type=int,
        default=128,
        help="integer to define the maximum length in tokens of each tune",
    )
    parser.add_argument(
        "-top_p",
        type=float,
        default=0.8,
        help="float to define the tokens that are within the sample operation of text generation",
    )
    parser.add_argument(
        "-top_k",
        type=int,
        default=8,
        help="integer to define the tokens that are within the sample operation of text generation",
    )
    parser.add_argument(
        "-temperature",
        type=float,
        default=1.2,
        help="the temperature of the sampling operation",
    )
    parser.add_argument("-seed", type=int, default=None, help="seed for randomstate")
    parser.add_argument(
        "-show_control_code",
        type=bool,
        default=False,
        help="whether to show control code",
    )
    return parser.parse_args()


def generate_music(args, region: str):
    patchilizer = Patchilizer()
    patch_config = GPT2Config(
        num_hidden_layers=PATCH_NUM_LAYERS,
        max_length=PATCH_LENGTH,
        max_position_embeddings=PATCH_LENGTH,
        vocab_size=1,
    )
    char_config = GPT2Config(
        num_hidden_layers=CHAR_NUM_LAYERS,
        max_length=PATCH_SIZE,
        max_position_embeddings=PATCH_SIZE,
        vocab_size=128,
    )
    model = TunesFormer(patch_config, char_config, share_weights=SHARE_WEIGHTS)
    checkpoint = torch.load(WEIGHTS_PATH, map_location=torch.device("cpu"))
    model.load_state_dict(checkpoint["model"])
    model = model.to(DEVICE)
    model.eval()
    prompt = f"A:{region}\n"
    tunes = ""
    num_tunes = args.num_tunes
    max_patch = args.max_patch
    top_p = args.top_p
    top_k = args.top_k
    temperature = args.temperature
    seed = args.seed
    show_control_code = args.show_control_code
    print(" Hyper parms ".center(60, "#"), "\n")
    arg_dict: dict = vars(args)
    for key in arg_dict.keys():
        print(f"{key}: {str(arg_dict[key])}")

    print("\n", " Output tunes ".center(60, "#"))
    start_time = time.time()
    for i in range(num_tunes):
        title_artist = f"T:{region} Style Fragment\nC:Generated by AI\n"
        tune = f"X:{str(i + 1)}\n{title_artist + prompt}"
        lines = re.split(r"(\n)", tune)
        tune = ""
        skip = False
        for line in lines:
            if show_control_code or line[:2] not in ["S:", "B:", "E:"]:
                if not skip:
                    print(line, end="")
                    tune += line

                skip = False

            else:
                skip = True

        input_patches = torch.tensor(
            [patchilizer.encode(prompt, add_special_patches=True)[:-1]], device=DEVICE
        )

        if tune == "":
            tokens = None

        else:
            prefix = patchilizer.decode(input_patches[0])
            remaining_tokens = prompt[len(prefix) :]
            tokens = torch.tensor(
                [patchilizer.bos_token_id] + [ord(c) for c in remaining_tokens],
                device=DEVICE,
            )

        while input_patches.shape[1] < max_patch:
            predicted_patch, seed = model.generate(
                input_patches,
                tokens,
                top_p=top_p,
                top_k=top_k,
                temperature=temperature,
                seed=seed,
            )
            tokens = None
            if predicted_patch[0] != patchilizer.eos_token_id:
                next_bar = patchilizer.decode([predicted_patch])
                if show_control_code or next_bar[:2] not in ["S:", "B:", "E:"]:
                    print(next_bar, end="")
                    tune += next_bar

                if next_bar == "":
                    break

                next_bar = remaining_tokens + next_bar
                remaining_tokens = ""
                predicted_patch = torch.tensor(
                    patchilizer.bar2patch(next_bar), device=DEVICE
                ).unsqueeze(0)
                input_patches = torch.cat(
                    [input_patches, predicted_patch.unsqueeze(0)], dim=1
                )

            else:
                break

        tunes += f"{tune}\n\n"
        print("\n")

    print("Generation time: {:.2f} seconds".format(time.time() - start_time))
    timestamp = time.strftime("%a_%d_%b_%Y_%H_%M_%S", time.localtime())
    try:
        xml = abc2xml(tunes, f"{TEMP_DIR}/[{region}]{timestamp}.musicxml")
        midi = xml2(xml, "mid")
        audio = xml2(xml, "wav")
        pdf, jpg = xml2img(xml)
        mxl = xml2(xml, "mxl")
        return audio, midi, pdf, xml, mxl, tunes, jpg

    except Exception as e:
        print(f"Invalid abc generated: {e}, retrying...")
        return generate_music(args, region)


def infer(p: float, k: int, t: float, region: str):
    status = "Success"
    audio = midi = pdf = xml = mxl = tunes = jpg = None
    try:
        if os.path.exists(TEMP_DIR):
            shutil.rmtree(TEMP_DIR)

        os.makedirs(TEMP_DIR)
        parser = argparse.ArgumentParser()
        args = get_args(parser)
        args.top_p = p
        args.top_k = k
        args.temperature = t
        audio, midi, pdf, xml, mxl, tunes, jpg = generate_music(
            args, region if EN_US else TEYVAT[region]
        )

    except Exception as e:
        status = f"{e}"

    return status, audio, midi, pdf, xml, mxl, tunes, jpg


if __name__ == "__main__":
    warnings.filterwarnings("ignore")
    opts = list(TEYVAT.values()) if EN_US else list(TEYVAT.keys())
    gr.Interface(
        fn=infer,
        inputs=[
            gr.Slider(0.01, 1.0, 0.8, step=0.01, label=_L("Top-P 采样")),
            gr.Slider(0, 80, 8, step=1, label=_L("Top-K 采样 (0 为关闭)")),
            gr.Slider(0.01, 2.0, 1.2, step=0.01, label=_L("温度参数")),
            gr.Dropdown(
                choices=opts,
                value=opts[0],
                label=_L("地区风格"),
            ),
        ],
        outputs=[
            gr.Textbox(label=_L("状态栏"), show_copy_button=True),
            gr.Audio(label=_L("音频"), type="filepath"),
            gr.File(label=_L("下载 MIDI")),
            gr.File(label=_L("下载 PDF 乐谱")),
            gr.File(label=_L("下载 MusicXML")),
            gr.File(label=_L("下载 MXL")),
            gr.Textbox(label=_L("ABC 记谱"), show_copy_button=True),
            gr.Image(label=_L("五线谱"), type="filepath", show_share_button=False),
        ],
        flagging_mode="never",
        title=_L("原神音乐生成"),
        description=_L(
            """
    欢迎使用此创空间, 此创空间基于 Tunesformer 开源项目制作，完全免费。当前模型还在调试中，计划在原神主线杀青后，所有国家地区角色全部开放后，二创音乐会齐全且样本均衡，届时重新微调模型并添加现实风格筛选辅助游戏各国家输出强化学习，以提升输出区分度与质量。注：崩铁方面数据工程正在运作中，未来也希望随主线杀青而基线化。<br>
    数据来源: <a href="https://musescore.org">MuseScore</a> 标签来源: <a href="https://genshin-impact.fandom.com/wiki/Genshin_Impact_Wiki">Genshin Impact Wiki | Fandom</a> 模型基础: <a href="https://github.com/sander-wood/tunesformer">Tunesformer</a>
    """
        ),
    ).launch(ssr_mode=False)