merge

Dockerfile.deploy

@@ -2,7 +2,7 @@
 
 # See https://huggingface.co/docs/hub/spaces-sdks-docker-first-demo
 
-FROM python:3.10
+FROM python:3.12
 
 RUN useradd -m -u 1000 user
 

Dockerfile.train

@@ -5,8 +5,8 @@
 
 # 主なバージョン等
 # Ubuntu 22.04
-# Python 3.10
-# PyTorch 2.1.2 (CUDA 11.8)
+# Python 3.12
+# PyTorch 2.7.x (CUDA 11.8)
 # CUDA Toolkit 12.0, CUDNN 8.9.7
 
 
@@ -24,7 +24,7 @@ ENV SHELL=/bin/bash
 ENV DEBIAN_FRONTEND=noninteractive
 
 ENV APT_INSTALL="apt-get install -y --no-install-recommends"
-ENV PIP_INSTALL="python3 -m pip --no-cache-dir install --upgrade"
+ENV PIP_INSTALL="python3.12 -m pip --no-cache-dir install --upgrade"
 ENV GIT_CLONE="git clone --depth 10"
 
 # ==================================================================
@@ -44,10 +44,14 @@ RUN apt-get update && \
     nano \
     ffmpeg \
     software-properties-common \
-    gnupg \
-    python3 \
-    python3-pip \
-    python3-dev
+    gnupg
+
+RUN add-apt-repository ppa:deadsnakes/ppa && \
+    apt-get update && \
+    $APT_INSTALL python3.12 python3.12-dev && \
+    curl -sS https://bootstrap.pypa.io/get-pip.py -o /tmp/get-pip.py && \
+    python3.12 /tmp/get-pip.py && \
+    rm /tmp/get-pip.py
 
 # ==================================================================
 # Git-lfs
@@ -57,8 +61,9 @@ RUN curl -s https://packagecloud.io/install/repositories/github/git-lfs/script.d
     $APT_INSTALL git-lfs
 
 
-# Add symlink so python and python3 commands use same python3.9 executable
-RUN ln -s /usr/bin/python3 /usr/local/bin/python    
+# Add symlinks so python and python3 commands use Python 3.12.
+RUN ln -s /usr/bin/python3.12 /usr/local/bin/python && \
+    ln -s /usr/bin/python3.12 /usr/local/bin/python3
 
 # ==================================================================
 # Installing CUDA packages (CUDA Toolkit 12.0 and CUDNN 8.9.7)

README.md

@@ -79,9 +79,9 @@ Pythonの仮想環境・パッケージ管理ツールである[uv](https://gith
 powershell -c "irm https://astral.sh/uv/install.ps1 | iex"
 git clone https://github.com/litagin02/Style-Bert-VITS2.git
 cd Style-Bert-VITS2
-uv venv venv
+uv venv --python 3.12 venv
 venv\Scripts\activate
-uv pip install "torch<2.4" "torchaudio<2.4" --index-url https://download.pytorch.org/whl/cu118
+uv pip install "torch>=2.8,<2.9" "torchaudio>=2.8,<2.9" --index-url https://download.pytorch.org/whl/cu126
 uv pip install -r requirements.txt
 python initialize.py  # 必要なモデルとデフォルトTTSモデルをダウンロード
 ```

app.py

@@ -49,7 +49,7 @@ model_holder = TTSModelHolder(
     ignore_onnx=True,
 )
 
-with gr.Blocks(theme=GRADIO_THEME) as app:
+with gr.Blocks() as app:
     gr.Markdown(f"# Style-Bert-VITS2 WebUI (version {VERSION})")
     with gr.Tabs():
         with gr.Tab("音声合成"):
@@ -66,6 +66,7 @@ with gr.Blocks(theme=GRADIO_THEME) as app:
             create_onnx_app(model_holder=model_holder)
 
 app.launch(
+    theme=GRADIO_THEME,
     server_name=args.host,
     server_port=args.port,
     inbrowser=not args.no_autolaunch,

bert_gen.py

@@ -55,7 +55,7 @@ def process_line(x: tuple[str, bool]):
     bert_path = wav_path.replace(".WAV", ".wav").replace(".wav", ".bert.pt")
 
     try:
-        bert = torch.load(bert_path)
+        bert = torch.load(bert_path, weights_only=True)
         assert bert.shape[-1] == len(phone)
     except Exception:
         bert = extract_bert_feature(text, word2ph, Languages(language_str), device)

data_utils.py

@@ -129,7 +129,7 @@ class TextAudioSpeakerLoader(torch.utils.data.Dataset):
         if self.use_mel_spec_posterior:
             spec_filename = spec_filename.replace(".spec.pt", ".mel.pt")
         try:
-            spec = torch.load(spec_filename)
+            spec = torch.load(spec_filename, weights_only=True)
         except:
             if self.use_mel_spec_posterior:
                 spec = mel_spectrogram_torch(
@@ -168,7 +168,7 @@ class TextAudioSpeakerLoader(torch.utils.data.Dataset):
             word2ph[0] += 1
         bert_path = wav_path.replace(".wav", ".bert.pt")
         try:
-            bert_ori = torch.load(bert_path)
+            bert_ori = torch.load(bert_path, weights_only=True)
             assert bert_ori.shape[-1] == len(phone)
         except Exception as e:
             logger.warning("Bert load Failed")

gradio_tabs/convert_onnx.py

@@ -52,7 +52,7 @@ def create_onnx_app(model_holder: TTSModelHolder) -> gr.Blocks:
     initial_id = 0
     initial_pth_files = get_model_files(model_names[initial_id])
 
-    with gr.Blocks(theme=GRADIO_THEME) as app:
+    with gr.Blocks() as app:
         gr.Markdown(initial_md)
         with gr.Row():
             with gr.Column():
@@ -100,4 +100,4 @@ if __name__ == "__main__":
     assets_root = path_config.assets_root
     model_holder = TTSModelHolder(assets_root, "cpu", "", ignore_onnx=True)
     app = create_onnx_app(model_holder)
-    app.launch(inbrowser=True)
+    app.launch(inbrowser=True, theme=GRADIO_THEME)

gradio_tabs/dataset.py

@@ -115,7 +115,7 @@ Style-Bert-VITS2の学習用データセットを作成するためのツール
 
 
 def create_dataset_app() -> gr.Blocks:
-    with gr.Blocks(theme=GRADIO_THEME) as app:
+    with gr.Blocks() as app:
         gr.Markdown(
             "**既に1ファイル2-12秒程度の音声ファイル集とその書き起こしデータがある場合は、このタブは使用せずに学習できます。**"
         )
@@ -273,4 +273,4 @@ def create_dataset_app() -> gr.Blocks:
 
 if __name__ == "__main__":
     app = create_dataset_app()
-    app.launch(inbrowser=True)
+    app.launch(inbrowser=True, theme=GRADIO_THEME)

gradio_tabs/inference.py

@@ -361,7 +361,7 @@ def create_inference_app(model_holder: TTSModelHolder) -> gr.Blocks:
     initial_id = 0
     initial_pth_files = get_model_files(model_names[initial_id])
 
-    with gr.Blocks(theme=GRADIO_THEME) as app:
+    with gr.Blocks() as app:
         gr.Markdown(initial_md)
         gr.Markdown(terms_of_use_md)
         null_models = gr.State({})
@@ -742,4 +742,4 @@ if __name__ == "__main__":
         assets_root, device, torch_device_to_onnx_providers(device)
     )
     app = create_inference_app(model_holder)
-    app.launch(inbrowser=True)
+    app.launch(inbrowser=True, theme=GRADIO_THEME)

gradio_tabs/merge.py

@@ -1042,7 +1042,7 @@ def create_merge_app(model_holder: TTSModelHolder) -> gr.Blocks:
         str(f) for f in model_holder.model_files_dict[model_names[initial_id]]
     ]
 
-    with gr.Blocks(theme=GRADIO_THEME) as app:
+    with gr.Blocks() as app:
         gr.Markdown(
             "複数のStyle-Bert-VITS2モデルから、声質・話し方・話す速さを取り替えたり混ぜたり引いたりして新しいモデルを作成できます。"
         )
@@ -1554,4 +1554,4 @@ if __name__ == "__main__":
         assets_root, device, torch_device_to_onnx_providers(device), ignore_onnx=True
     )
     app = create_merge_app(model_holder)
-    app.launch(inbrowser=True)
+    app.launch(inbrowser=True, theme=GRADIO_THEME)

gradio_tabs/style_vectors.py

@@ -416,7 +416,7 @@ https://ja.wikipedia.org/wiki/DBSCAN
 
 
 def create_style_vectors_app():
-    with gr.Blocks(theme=GRADIO_THEME) as app:
+    with gr.Blocks() as app:
         with gr.Accordion("使い方", open=False):
             gr.Markdown(how_to_md)
         model_name = gr.Textbox(placeholder="your_model_name", label="モデル名")
@@ -583,4 +583,4 @@ def create_style_vectors_app():
 
 if __name__ == "__main__":
     app = create_style_vectors_app()
-    app.launch(inbrowser=True)
+    app.launch(inbrowser=True, theme=GRADIO_THEME)

gradio_tabs/train.py

@@ -492,7 +492,7 @@ english_teacher.wav|Mary|EN|How are you? I'm fine, thank you, and you?
 
 
 def create_train_app():
-    with gr.Blocks(theme=GRADIO_THEME).queue() as app:
+    with gr.Blocks().queue() as app:
         gr.Markdown(change_log_md)
         with gr.Accordion("使い方", open=False):
             gr.Markdown(how_to_md)
@@ -852,4 +852,4 @@ def create_train_app():
 
 if __name__ == "__main__":
     app = create_train_app()
-    app.launch(inbrowser=True)
+    app.launch(inbrowser=True, theme=GRADIO_THEME)

pyproject.toml

@@ -7,7 +7,7 @@ name = "style-bert-vits2"
 dynamic = ["version"]
 description = "Style-Bert-VITS2: Bert-VITS2 with more controllable voice styles."
 readme = "README.md"
-requires-python = ">=3.9"
+requires-python = ">=3.12"
 license = "AGPL-3.0"
 keywords = []
 authors = [
@@ -16,9 +16,7 @@ authors = [
 classifiers = [
     "Development Status :: 4 - Beta",
     "Programming Language :: Python",
-    "Programming Language :: Python :: 3.9",
-    "Programming Language :: Python :: 3.10",
-    "Programming Language :: Python :: 3.11",
+    "Programming Language :: Python :: 3.12",
     "Programming Language :: Python :: Implementation :: CPython",
 ]
 dependencies = [
@@ -27,13 +25,13 @@ dependencies = [
     "g2p_en",
     "jieba",
     "loguru",
-    "nltk<=3.8.1",
+    "nltk>=3.9,<4",
     "num2words",
     "numba",
-    "numpy<2",
+    "numpy>=2,<2.5",
     "onnxruntime",
     "pydantic>=2.0",
-    "pyopenjtalk-dict",
+    "pyopenjtalk-plus",
     "pypinyin",
     "pyworld-prebuilt",
     "safetensors",
@@ -43,7 +41,7 @@ dependencies = [
 [project.optional-dependencies]
 torch = [
     "accelerate",
-    "torch>=2.1",
+    "torch>=2.8,<2.9",
 ]
 
 [project.urls]
@@ -93,7 +91,7 @@ cov-report = ["- coverage combine", "coverage report"]
 # Usage: `hatch run test:cov`
 cov = ["test-cov", "cov-report"]
 [[tool.hatch.envs.test.matrix]]
-python = ["3.9", "3.10", "3.11"]
+python = ["3.12"]
 
 # for ONNX inference (without PyTorch dependency)
 [tool.hatch.envs.test-onnx]
@@ -120,7 +118,7 @@ cov-report = ["- coverage combine", "coverage report"]
 # Usage: `hatch run test-onnx:cov`
 cov = ["test-cov", "cov-report"]
 [[tool.hatch.envs.test-onnx.matrix]]
-python = ["3.9", "3.10", "3.11"]
+python = ["3.12"]
 
 [tool.hatch.envs.style]
 detached = true

requirements-colab.txt

@@ -4,19 +4,19 @@ cn2an
 g2p_en
 gradio>=4.32
 jieba
-librosa==0.9.2
+librosa>=0.11,<1
 loguru
-nltk<=3.8.1
+nltk>=3.9,<4
 num2words
-numpy<2
+numpy>=2,<2.5
 onnx
 onnxconverter-common
 onnxruntime
 onnxruntime-gpu
 onnxsim-prebuilt
-pyannote.audio>=3.1.0
+pyannote.audio>=4,<5
 pyloudnorm
-pyopenjtalk-dict
+pyopenjtalk-plus
 pypinyin
 pyworld-prebuilt
 torch

requirements-infer.txt

@@ -1,33 +1,33 @@
 accelerate
 cmudict
 cn2an
-# faster-whisper==0.10.1
+# faster-whisper>=1.2,<2
 g2p_en
 GPUtil
 gradio>=4.32
 jieba
-# librosa==0.9.2
+# librosa>=0.11,<1
 loguru
-nltk<=3.8.1
+nltk>=3.9,<4
 num2words
-numpy<2
+numpy>=2,<2.5
 onnx
 onnxconverter-common
 onnxruntime
 onnxruntime-directml; sys_platform == 'win32'
 onnxruntime-gpu; sys_platform != 'darwin'
 onnxsim-prebuilt
-# protobuf==4.25
+# protobuf>=5,<7
 psutil
 # punctuators
-pyannote.audio>=3.1.0
+pyannote.audio>=4,<5
 # pyloudnorm
-pyopenjtalk-dict
+pyopenjtalk-plus
 pypinyin
 pyworld-prebuilt
 # stable_ts
 # tensorboard
-torch<2.4
-torchaudio<2.4
+torch>=2.8,<2.9
+torchaudio>=2.8,<2.9
 transformers
 umap-learn

requirements.txt

@@ -1,33 +1,33 @@
 accelerate
 cmudict
 cn2an
-faster-whisper==0.10.1
+faster-whisper>=1.2,<2
 g2p_en
 GPUtil
 gradio>=4.32
 jieba
-librosa==0.9.2
+librosa>=0.11,<1
 loguru
-nltk<=3.8.1
+nltk>=3.9,<4
 num2words
-numpy<2
+numpy>=2,<2.5
 onnx
 onnxconverter-common
 onnxruntime
 onnxruntime-directml; sys_platform == 'win32'
 onnxruntime-gpu; sys_platform != 'darwin'
 onnxsim-prebuilt
-protobuf==4.25
+protobuf>=5,<7
 psutil
 punctuators
-pyannote.audio>=3.1.0
+pyannote.audio>=4,<5
 pyloudnorm
-pyopenjtalk-dict
+pyopenjtalk-plus
 pypinyin
 pyworld-prebuilt
 stable_ts
 tensorboard
-torch<2.4
-torchaudio<2.4
+torch>=2.8,<2.9
+torchaudio>=2.8,<2.9
 transformers
 umap-learn

scripts/Install-Style-Bert-VITS2.bat

@@ -110,8 +110,8 @@ if !errorlevel! neq 0 ( pause & popd & exit /b !errorlevel! )
 echo --------------------------------------------------
 echo Installing PyTorch...
 echo --------------------------------------------------
-echo Executing: uv pip install "torch<2.4" "torchaudio<2.4" --index-url https://download.pytorch.org/whl/cu118
-uv pip install "torch<2.4" "torchaudio<2.4" --index-url https://download.pytorch.org/whl/cu118
+echo Executing: uv pip install "torch>=2.8,<2.9" "torchaudio>=2.8,<2.9" --index-url https://download.pytorch.org/whl/cu126
+uv pip install "torch>=2.8,<2.9" "torchaudio>=2.8,<2.9" --index-url https://download.pytorch.org/whl/cu126
 if !errorlevel! neq 0 ( pause & popd & exit /b !errorlevel! )
 
 echo --------------------------------------------------

scripts/Setup-Python.bat

@@ -40,8 +40,8 @@ if not exist "%PYTHON_DIR%"\ (
 	echo --------------------------------------------------
 	echo Downloading Python...
 	echo --------------------------------------------------
-	echo Executing: %CURL_CMD% -o python.zip https://www.python.org/ftp/python/3.10.11/python-3.10.11-embed-amd64.zip
-	%CURL_CMD% -o python.zip https://www.python.org/ftp/python/3.10.11/python-3.10.11-embed-amd64.zip
+	echo Executing: %CURL_CMD% -o python.zip https://www.python.org/ftp/python/3.12.10/python-3.12.10-embed-amd64.zip
+	%CURL_CMD% -o python.zip https://www.python.org/ftp/python/3.12.10/python-3.12.10-embed-amd64.zip
 	if !errorlevel! neq 0 ( pause & exit /b !errorlevel! )
 
 	echo --------------------------------------------------
@@ -61,8 +61,8 @@ if not exist "%PYTHON_DIR%"\ (
 	echo --------------------------------------------------
 	echo Enabling 'site' module in the embedded Python environment...
 	echo --------------------------------------------------
-	echo Executing: %PS_CMD% "&{(Get-Content '%PYTHON_DIR%/python310._pth') -creplace '#import site', 'import site' | Set-Content '%PYTHON_DIR%/python310._pth' }"
-	%PS_CMD% "&{(Get-Content '%PYTHON_DIR%/python310._pth') -creplace '#import site', 'import site' | Set-Content '%PYTHON_DIR%/python310._pth' }"
+	echo Executing: %PS_CMD% "&{(Get-Content '%PYTHON_DIR%/python312._pth') -creplace '#import site', 'import site' | Set-Content '%PYTHON_DIR%/python312._pth' }"
+	%PS_CMD% "&{(Get-Content '%PYTHON_DIR%/python312._pth') -creplace '#import site', 'import site' | Set-Content '%PYTHON_DIR%/python312._pth' }"
 	if !errorlevel! neq 0 ( pause & exit /b !errorlevel! )
 
 	echo --------------------------------------------------

slice.py

@@ -171,7 +171,7 @@ if __name__ == "__main__":
 
     # モデルをダウンロードしておく
     _ = torch.hub.load(
-        repo_or_dir="litagin02/silero-vad",
+        repo_or_dir="snakers4/silero-vad",
         model="silero_vad",
         onnx=True,
         trust_repo=True,
@@ -186,7 +186,7 @@ if __name__ == "__main__":
     ):
         # logger.debug("Worker started.")
         vad_model, utils = torch.hub.load(
-            repo_or_dir="litagin02/silero-vad",
+            repo_or_dir="snakers4/silero-vad",
             model="silero_vad",
             onnx=True,
             trust_repo=True,

style_bert_vits2/models/utils/checkpoints.py

@@ -32,7 +32,9 @@ def load_checkpoint(
     """
 
     assert os.path.isfile(checkpoint_path)
-    checkpoint_dict = torch.load(checkpoint_path, map_location=device)
+    checkpoint_dict = torch.load(
+        checkpoint_path, map_location=device, weights_only=True
+    )
     iteration = checkpoint_dict["iteration"]
     learning_rate = checkpoint_dict["learning_rate"]
     logger.info(

style_bert_vits2/nlp/japanese/user_dict/word_model.py

@@ -9,7 +9,13 @@ from enum import Enum
 from re import findall, fullmatch
 from typing import List, Optional
 
-from pydantic import BaseModel, Field, validator
+from pydantic import (
+    BaseModel,
+    ConfigDict,
+    Field,
+    ValidationInfo,
+    field_validator,
+)
 
 
 USER_DICT_MIN_PRIORITY = 0
@@ -39,11 +45,11 @@ class UserDictWord(BaseModel):
     mora_count: Optional[int] = Field(title="モーラ数", default=None)
     accent_associative_rule: str = Field(title="アクセント結合規則")
 
-    class Config:
-        validate_assignment = True
+    model_config = ConfigDict(validate_assignment=True, validate_default=True)
 
-    @validator("surface")
-    def convert_to_zenkaku(cls, surface):
+    @field_validator("surface")
+    @classmethod
+    def convert_to_zenkaku(cls, surface: str) -> str:
         return surface.translate(
             str.maketrans(
                 "".join(chr(0x21 + i) for i in range(94)),
@@ -51,8 +57,9 @@ class UserDictWord(BaseModel):
             )
         )
 
-    @validator("pronunciation", pre=True)
-    def check_is_katakana(cls, pronunciation):
+    @field_validator("pronunciation", mode="before")
+    @classmethod
+    def check_is_katakana(cls, pronunciation: str) -> str:
         if not fullmatch(r"[ァ-ヴー]+", pronunciation):
             raise ValueError("発音は有効なカタカナでなくてはいけません。")
         sutegana = ["ァ", "ィ", "ゥ", "ェ", "ォ", "ャ", "ュ", "ョ", "ヮ", "ッ"]
@@ -75,9 +82,12 @@ class UserDictWord(BaseModel):
                     )
         return pronunciation
 
-    @validator("mora_count", pre=True, always=True)
-    def check_mora_count_and_accent_type(cls, mora_count, values):
-        if "pronunciation" not in values or "accent_type" not in values:
+    @field_validator("mora_count", mode="before")
+    @classmethod
+    def check_mora_count_and_accent_type(
+        cls, mora_count: Optional[int], info: ValidationInfo
+    ) -> Optional[int]:
+        if "pronunciation" not in info.data or "accent_type" not in info.data:
             # 適切な場所でエラーを出すようにする
             return mora_count
 
@@ -91,14 +101,14 @@ class UserDictWord(BaseModel):
             mora_count = len(
                 findall(
                     f"(?:{rule_others}|{rule_line_i}|{rule_line_u}|{rule_one_mora})",
-                    values["pronunciation"],
+                    info.data["pronunciation"],
                 )
             )
 
-        if not 0 <= values["accent_type"] <= mora_count:
+        if not 0 <= info.data["accent_type"] <= mora_count:
             raise ValueError(
                 "誤ったアクセント型です({})。 expect: 0 <= accent_type <= {}".format(
-                    values["accent_type"], mora_count
+                    info.data["accent_type"], mora_count
                 )
             )
         return mora_count

tests/test_library_compatibility.py

@@ -0,0 +1,58 @@
+from pathlib import Path
+
+import pytest
+import torch
+from pydantic import ValidationError
+
+from style_bert_vits2.models.utils.checkpoints import load_checkpoint, save_checkpoint
+from style_bert_vits2.nlp.japanese.user_dict.word_model import UserDictWord
+
+
+def _user_dict_word(**overrides: object) -> UserDictWord:
+    values = {
+        "surface": "test",
+        "priority": 5,
+        "part_of_speech": "名詞",
+        "part_of_speech_detail_1": "一般",
+        "part_of_speech_detail_2": "*",
+        "part_of_speech_detail_3": "*",
+        "inflectional_type": "*",
+        "inflectional_form": "*",
+        "stem": "*",
+        "yomi": "テスト",
+        "pronunciation": "テスト",
+        "accent_type": 1,
+        "accent_associative_rule": "*",
+    }
+    values.update(overrides)
+    return UserDictWord.model_validate(values)
+
+
+def test_user_dict_word_uses_pydantic_v2_validation() -> None:
+    word = _user_dict_word()
+
+    assert word.surface == "ｔｅｓｔ"
+    assert word.mora_count == 3
+
+    with pytest.raises(ValidationError):
+        _user_dict_word(pronunciation="てすと")
+
+
+def test_checkpoint_round_trip_with_safe_torch_load(tmp_path: Path) -> None:
+    checkpoint_path = tmp_path / "checkpoint.pth"
+    source = torch.nn.Linear(2, 1)
+    optimizer = torch.optim.AdamW(source.parameters())
+    save_checkpoint(source, optimizer, 1e-4, 12, checkpoint_path)
+
+    destination = torch.nn.Linear(2, 1)
+    destination_optimizer = torch.optim.AdamW(destination.parameters())
+    _, _, learning_rate, iteration = load_checkpoint(
+        checkpoint_path, destination, destination_optimizer
+    )
+
+    assert learning_rate == 1e-4
+    assert iteration == 12
+    for source_parameter, destination_parameter in zip(
+        source.parameters(), destination.parameters()
+    ):
+        assert torch.equal(source_parameter, destination_parameter)