chore(format): run black on dev

Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>

.github/workflows/checksum.yml

@@ -15,7 +15,7 @@ jobs:
 
       - name: Run RVC-Models-Downloader
         run: |
-          wget https://github.com/fumiama/RVC-Models-Downloader/releases/download/v0.2.5/rvcmd_linux_amd64.deb
+          wget https://github.com/fumiama/RVC-Models-Downloader/releases/download/v0.2.11/rvcmd_linux_amd64.deb
           sudo apt -y install ./rvcmd_linux_amd64.deb
           rm -f ./rvcmd_linux_amd64.deb
           rvcmd -notrs -w 1 -notui assets/rvc

.github/workflows/close-issue.yml

@@ -12,7 +12,7 @@ jobs:
     steps:
       - uses: actions/stale@v5
         with:
-          exempt-issue-labels: "help wanted,good first issue,documentation,following up,todo list"
+          exempt-issue-labels: "help wanted,good first issue,documentation,following up,todo list,enhancement"
           days-before-issue-stale: 30
           days-before-issue-close: 15
           stale-issue-label: "stale"

.github/workflows/unitest.yml

@@ -11,20 +11,72 @@ jobs:
 
     steps:
     - uses: actions/checkout@master
+    - name: Space cleanup
+      env:
+        DEBIAN_FRONTEND: noninteractive
+      run: |
+        df -h
+
+        # Source - https://stackoverflow.com/a
+        # Posted by Cosmin Bodnariuc
+        # Retrieved 2025-11-21, License - CC BY-SA 4.0
+        sudo rm -rf /usr/share/dotnet
+        sudo rm -rf /usr/local/share/boost
+        sudo rm -rf /usr/local/share/chromium
+        sudo rm -rf /usr/local/share/powershell
+        sudo rm -rf /usr/local/share/vcpkg
+        sudo rm -rf /usr/local/share/miniconda
+        sudo rm -rf /opt/ghc
+        sudo rm -rf /opt/hostedtoolcache/CodeQL
+        sudo rm -rf /opt/hostedtoolcache/go
+        sudo rm -rf /opt/hostedtoolcache/Python
+        sudo rm -rf /opt/hostedtoolcache/node
+        sudo rm -rf /opt/hostedtoolcache/R
+        sudo rm -rf /opt/hostedtoolcache/Java
+        sudo rm -rf /opt/hostedtoolcache/LLVM
+        sudo rm -rf /opt/hostedtoolcache/Swift
+        sudo rm -rf /opt/hostedtoolcache/Php
+        sudo rm -rf /opt/hostedtoolcache/Perl
+        sudo rm -rf /opt/hostedtoolcache/Scala
+        sudo rm -rf /opt/hostedtoolcache/Julia
+        sudo rm -rf /opt/hostedtoolcache/Mono
+        sudo rm -rf /opt/hostedtoolcache/PowerShell
+        sudo rm -rf /opt/hostedtoolcache/Crystal
+        sudo rm -rf /opt/hostedtoolcache/Elixir
+        sudo rm -rf /opt/hostedtoolcache/Erlang
+        sudo rm -rf /opt/hostedtoolcache/FSharp
+        sudo rm -rf /opt/hostedtoolcache/Haskell
+        sudo rm -rf /opt/hostedtoolcache/OCaml
+        sudo rm -rf /opt/hostedtoolcache/Rust
+        sudo rm -rf /opt/hostedtoolcache/Sbt
+        sudo rm -rf /opt/hostedtoolcache/Solidity
+        sudo rm -rf /opt/hostedtoolcache/VisualStudio
+        sudo rm -rf /opt/hostedtoolcache/WinAppDriver
+        sudo rm -rf /opt/hostedtoolcache/Xamarin
+        sudo rm -rf /opt/hostedtoolcache/Yarn
+        sudo rm -rf /opt/hostedtoolcache/Zephyr
+        sudo rm -rf /opt/hostedtoolcache/zig
+        sudo rm -rf /opt/hostedtoolcache/zulu
+        sudo rm -rf /opt/hostedtoolcache/azcopy  
+
+        sudo -E apt-get update
+        sudo -E apt-get -y autoremove --purge
+        sudo -E apt-get clean
+
+        df -h
     - name: Set up Python ${{ matrix.python-version }}
       uses: actions/setup-python@v4
       with:
         python-version: ${{ matrix.python-version }}
     - name: Install dependencies
       run: |
-        sudo apt update
-        wget https://github.com/fumiama/RVC-Models-Downloader/releases/download/v0.2.5/rvcmd_linux_amd64.deb
+        wget https://github.com/fumiama/RVC-Models-Downloader/releases/download/v0.2.11/rvcmd_linux_amd64.deb
         sudo apt -y install ./rvcmd_linux_amd64.deb
         pip install --force pip==24.0 # fix fairseq installing issue https://github.com/facebookresearch/fairseq/issues/5552
         python -m pip install --upgrade setuptools
         python -m pip install --upgrade wheel
         pip install torch torchvision torchaudio
-        pip install -r requirements/main.txt
+        pip install -r requirements/cpu.txt
         rvcmd -notrs -w 1 -notui assets/rvc
     - name: Test step 1 & 2
       run: |

configs/config.py

@@ -4,6 +4,7 @@ import sys
 import json
 import shutil
 from multiprocessing import cpu_count
+import importlib.util
 
 import torch
 
@@ -46,10 +47,10 @@ class Config(metaclass=Singleton):
             self.global_link,
             self.noparallel,
             self.noautoopen,
-            self.dml,
             self.nocheck,
             self.update,
         ) = self.arg_parse()
+        self.dml = False
         self.instead = ""
         self.preprocess_per = 3.7
         self.x_pad, self.x_query, self.x_center, self.x_max = self.device_config()
@@ -83,11 +84,6 @@ class Config(metaclass=Singleton):
             action="store_true",
             help="Do not open in browser automatically",
         )
-        parser.add_argument(
-            "--dml",
-            action="store_true",
-            help="torch_dml",
-        )
         parser.add_argument(
             "--nocheck", action="store_true", help="Run without checking assets"
         )
@@ -104,7 +100,6 @@ class Config(metaclass=Singleton):
             cmd_opts.global_link,
             cmd_opts.noparallel,
             cmd_opts.noautoopen,
-            cmd_opts.dml,
             cmd_opts.nocheck,
             cmd_opts.update,
         )
@@ -183,7 +178,7 @@ class Config(metaclass=Singleton):
             if self.has_xpu():
                 self.device = self.instead = "xpu:0"
                 self.is_half = True
-            i_device = int(self.device.split(":")[-1])
+            i_device = int(str(self.device).split(":")[-1])
             self.gpu_name = torch.cuda.get_device_name(i_device)
             if (
                 ("16" in self.gpu_name and "V100" not in self.gpu_name.upper())
@@ -214,7 +209,7 @@ class Config(metaclass=Singleton):
             self.use_fp32_config()
         else:
             logger.info("No supported Nvidia GPU found")
-            self.device = self.instead = "cpu"
+            self.device = self.instead = torch.get_default_device()
             self.is_half = False
             self.use_fp32_config()
 
@@ -239,12 +234,13 @@ class Config(metaclass=Singleton):
             x_query = 5
             x_center = 30
             x_max = 32
-        if self.dml:
+        if importlib.util.find_spec("torch_directml") is not None:
             logger.info("Use DirectML instead")
             import torch_directml
 
             self.device = torch_directml.device(torch_directml.default_device())
             self.is_half = False
+            self.dml = True
         else:
             if self.instead:
                 logger.info(f"Use {self.instead} instead")

i18n/locale/en_US.json

@@ -140,7 +140,6 @@
     "Train model": "Train model",
     "Training complete. You can check the training logs in the console or the 'train.log' file under the experiment folder.": "Training complete. You can check the training logs in the console or the 'train.log' file under the experiment folder.",
     "Transpose (integer, number of semitones, raise by an octave: 12, lower by an octave: -12)": "Transpose (integer, number of semitones, raise by an octave: 12, lower by an octave: -12)",
-    "Unfortunately, there is no compatible GPU available to support your training.": "Unfortunately, there is no compatible GPU available to support your training.",
     "Unknown": "Unknown",
     "Unload model to save GPU memory": "Unload model to save GPU memory",
     "Version": "Version",

i18n/locale/es_ES.json

@@ -140,7 +140,6 @@
     "Train model": "Entrenar Modelo",
     "Training complete. You can check the training logs in the console or the 'train.log' file under the experiment folder.": "Entrenamiento finalizado, puede ver el registro de entrenamiento en la consola o en el archivo train.log en la carpeta del experimento",
     "Transpose (integer, number of semitones, raise by an octave: 12, lower by an octave: -12)": "Cambio de tono (entero, número de semitonos, subir una octava +12 o bajar una octava -12)",
-    "Unfortunately, there is no compatible GPU available to support your training.": "Lamentablemente, no tiene una tarjeta gráfica adecuada para soportar su entrenamiento",
     "Unknown": "Desconocido",
     "Unload model to save GPU memory": "Descargue la voz para ahorrar memoria GPU",
     "Version": "Versión",

i18n/locale/fr_FR.json

@@ -140,7 +140,6 @@
     "Train model": "Entraîner le modèle",
     "Training complete. You can check the training logs in the console or the 'train.log' file under the experiment folder.": "Entraînement terminé. Vous pouvez consulter les rapports d'entraînement dans la console ou dans le fichier 'train.log' situé dans le dossier de l'expérience.",
     "Transpose (integer, number of semitones, raise by an octave: 12, lower by an octave: -12)": "Transposer (entier, nombre de demi-tons, monter d'une octave : 12, descendre d'une octave : -12) :",
-    "Unfortunately, there is no compatible GPU available to support your training.": "Malheureusement, il n'y a pas de GPU compatible disponible pour prendre en charge votre entrainement.",
     "Unknown": "Inconnu",
     "Unload model to save GPU memory": "Décharger la voix pour économiser la mémoire GPU.",
     "Version": "Version",

i18n/locale/it_IT.json

@@ -140,7 +140,6 @@
     "Train model": "Addestra modello",
     "Training complete. You can check the training logs in the console or the 'train.log' file under the experiment folder.": "Addestramento completato. ",
     "Transpose (integer, number of semitones, raise by an octave: 12, lower by an octave: -12)": "Trasposizione (numero intero, numero di semitoni, alza di un'ottava: 12, abbassa di un'ottava: -12):",
-    "Unfortunately, there is no compatible GPU available to support your training.": "Sfortunatamente, non è disponibile alcuna GPU compatibile per supportare l'addestramento.",
     "Unknown": "Unknown",
     "Unload model to save GPU memory": "Scarica la voce per risparmiare memoria della GPU:",
     "Version": "Versione",

i18n/locale/ja_JP.json

@@ -133,14 +133,13 @@
     "Takeover WASAPI device": "WASAPIデバイスを独占",
     "Target sample rate": "目標サンプリング率",
     "The audio file to be processed": "処理待ち音声",
-    "This software is licensed under the GNU Affero General Public License, version 3.0 or later.<br>The author has no control or responsibility regarding the use of this software.<br>Users who use the software and distribute any content, including sounds or files generated by it, bear full responsibility for compliance with the terms of the AGPL 3.0 license.<br>If you do not accept these terms, you are prohibited from using, referencing, or distributing any code or files contained within this software package.<br>Refer to the LICENSE file located in the root directory for full details.": "This software is licensed under the GNU Affero General Public License, version 3.0 or later.<br>The author has no control or responsibility regarding the use of this software.<br>Users who use the software and distribute any content, including sounds or files generated by it, bear full responsibility for compliance with the terms of the AGPL 3.0 license.<br>If you do not accept these terms, you are prohibited from using, referencing, or distributing any code or files contained within this software package.<br>Refer to the LICENSE file located in the root directory for full details.",
+    "This software is licensed under the GNU Affero General Public License, version 3.0 or later.<br>The author has no control or responsibility regarding the use of this software.<br>Users who use the software and distribute any content, including sounds or files generated by it, bear full responsibility for compliance with the terms of the AGPL 3.0 license.<br>If you do not accept these terms, you are prohibited from using, referencing, or distributing any code or files contained within this software package.<br>Refer to the LICENSE file located in the root directory for full details.": "本ソフトウェアは、GNU Affero General Public License バージョン3.0以降の下でライセンスされています。<br>作者は、このソフトウェアの使用に関して一切の管理権限や責任を負いません。<br>このソフトウェアを使用し、生成された音声やファイルを含むあらゆるコンテンツを配布するユーザーは、AGPL 3.0ライセンスの条項への準拠について全責任を負います。<br>これらの条項に同意しない場合、このソフトウェアパッケージに含まれるコードやファイルの使用、参照、配布は禁止されています。<br>詳細については、ルートディレクトリにあるLICENSEファイルを参照してください。",
     "Total training epochs (total_epoch)": "総エポック数",
     "Train": "学習",
     "Train feature index": "特徴索引の学習",
     "Train model": "モデルの学習",
     "Training complete. You can check the training logs in the console or the 'train.log' file under the experiment folder.": "学習終了時に、学習ログやフォルダ内のtrain.logを確認することができます",
     "Transpose (integer, number of semitones, raise by an octave: 12, lower by an octave: -12)": "ピッチ変更(整数、半音数、上下オクターブ12-12)",
-    "Unfortunately, there is no compatible GPU available to support your training.": "学習に対応したGPUが動作しないのは残念です。",
     "Unknown": "未知",
     "Unload model to save GPU memory": "音源を削除してメモリを節約",
     "Version": "バージョン",

i18n/locale/ko_KR.json

@@ -140,7 +140,6 @@
     "Train model": "모델 훈련",
     "Training complete. You can check the training logs in the console or the 'train.log' file under the experiment folder.": "훈련 완료, 콘솔 훈련 로그 또는 실험 폴더 내의 train.log 확인 가능",
     "Transpose (integer, number of semitones, raise by an octave: 12, lower by an octave: -12)": "키 변경(정수, 반음 수, 옥타브 상승 12, 옥타브 하강 -12)",
-    "Unfortunately, there is no compatible GPU available to support your training.": "사용 가능한 그래픽 카드가 없어 훈련을 지원할 수 없습니다",
     "Unknown": "Unknown",
     "Unload model to save GPU memory": "음색 언로드로 디스플레이 메모리 절약",
     "Version": "버전",

i18n/locale/pt_BR.json

@@ -140,7 +140,6 @@
     "Train model": "Treinar Modelo",
     "Training complete. You can check the training logs in the console or the 'train.log' file under the experiment folder.": "Após o término do treinamento, você pode verificar o log de treinamento do console ou train.log na pasta de experimentos",
     "Transpose (integer, number of semitones, raise by an octave: 12, lower by an octave: -12)": "Mude o tom aqui. Se a voz for do mesmo sexo, não é necessario alterar (12 caso seja Masculino para feminino, -12 caso seja ao contrário).",
-    "Unfortunately, there is no compatible GPU available to support your training.": "Infelizmente, não há GPU compatível disponível para apoiar o seu treinamento.",
     "Unknown": "Unknown",
     "Unload model to save GPU memory": "Descarregue a voz para liberar a memória da GPU:",
     "Version": "Versão",

i18n/locale/ru_RU.json

@@ -140,7 +140,6 @@
     "Train model": "Обучить модель",
     "Training complete. You can check the training logs in the console or the 'train.log' file under the experiment folder.": "Обучение модели завершено. Журнал обучения можно просмотреть в консоли или в файле 'train.log' в папке с моделью.",
     "Transpose (integer, number of semitones, raise by an octave: 12, lower by an octave: -12)": "Изменить высоту голоса (укажите количество полутонов; чтобы поднять голос на октаву, выберите 12, понизить на октаву — -12):",
-    "Unfortunately, there is no compatible GPU available to support your training.": "К сожалению, у вас нету графического процессора, который поддерживает обучение моделей.",
     "Unknown": "Unknown",
     "Unload model to save GPU memory": "Выгрузить модель из памяти GPU для освобождения ресурсов",
     "Version": "Версия архитектуры модели:",

i18n/locale/tr_TR.json

@@ -140,7 +140,6 @@
     "Train model": "Modeli Eğit",
     "Training complete. You can check the training logs in the console or the 'train.log' file under the experiment folder.": "Eğitim tamamlandı. Eğitim günlüklerini konsolda veya deney klasörü altındaki train.log dosyasında kontrol edebilirsiniz.",
     "Transpose (integer, number of semitones, raise by an octave: 12, lower by an octave: -12)": "Transpoze et (tamsayı, yarıton sayısıyla; bir oktav yükseltmek için: 12, bir oktav düşürmek için: -12):",
-    "Unfortunately, there is no compatible GPU available to support your training.": "Maalesef, eğitiminizi desteklemek için uyumlu bir GPU bulunmamaktadır.",
     "Unknown": "Unknown",
     "Unload model to save GPU memory": "GPU bellek kullanımını azaltmak için sesi kaldır",
     "Version": "Sürüm",

i18n/locale/zh_CN.json

@@ -133,14 +133,13 @@
     "Takeover WASAPI device": "独占 WASAPI 设备",
     "Target sample rate": "目标采样率",
     "The audio file to be processed": "待处理音频文件",
-    "This software is licensed under the GNU Affero General Public License, version 3.0 or later.<br>The author has no control or responsibility regarding the use of this software.<br>Users who use the software and distribute any content, including sounds or files generated by it, bear full responsibility for compliance with the terms of the AGPL 3.0 license.<br>If you do not accept these terms, you are prohibited from using, referencing, or distributing any code or files contained within this software package.<br>Refer to the LICENSE file located in the root directory for full details.": "This software is licensed under the GNU Affero General Public License, version 3.0 or later.<br>The author has no control or responsibility regarding the use of this software.<br>Users who use the software and distribute any content, including sounds or files generated by it, bear full responsibility for compliance with the terms of the AGPL 3.0 license.<br>If you do not accept these terms, you are prohibited from using, referencing, or distributing any code or files contained within this software package.<br>Refer to the LICENSE file located in the root directory for full details.",
+    "This software is licensed under the GNU Affero General Public License, version 3.0 or later.<br>The author has no control or responsibility regarding the use of this software.<br>Users who use the software and distribute any content, including sounds or files generated by it, bear full responsibility for compliance with the terms of the AGPL 3.0 license.<br>If you do not accept these terms, you are prohibited from using, referencing, or distributing any code or files contained within this software package.<br>Refer to the LICENSE file located in the root directory for full details.": "本软件基于GNU Affero通用公共许可证3.0版本或更高版本进行许可。<br>作者对本软件的使用不承担任何控制权或责任。<br>使用本软件并分发任何内容（包括由其生成的声音或文件）的用户，需对遵守AGPL 3.0许可证条款承担全部责任。<br>如果您不接受这些条款，则禁止使用、引用或分发本软件包中包含的任何代码或文件。<br>请参阅位于根目录中的LICENSE文件以获取完整详情。",
     "Total training epochs (total_epoch)": "总训练轮数total_epoch",
     "Train": "训练",
     "Train feature index": "训练特征索引",
     "Train model": "训练模型",
     "Training complete. You can check the training logs in the console or the 'train.log' file under the experiment folder.": "训练结束, 您可查看控制台训练日志或实验文件夹下的train.log",
     "Transpose (integer, number of semitones, raise by an octave: 12, lower by an octave: -12)": "变调(整数, 半音数量, 升八度12降八度-12)",
-    "Unfortunately, there is no compatible GPU available to support your training.": "很遗憾您这没有能用的显卡来支持您训练",
     "Unknown": "未知",
     "Unload model to save GPU memory": "卸载音色省显存",
     "Version": "版本",

i18n/locale/zh_HK.json

@@ -140,7 +140,6 @@
     "Train model": "訓練模型",
     "Training complete. You can check the training logs in the console or the 'train.log' file under the experiment folder.": "训练结束, 您可查看控制台训练日志或实验文件夹下的train.log",
     "Transpose (integer, number of semitones, raise by an octave: 12, lower by an octave: -12)": "變調(整數、半音數量、升八度12降八度-12)",
-    "Unfortunately, there is no compatible GPU available to support your training.": "很遗憾您这没有能用的显卡来支持您训练",
     "Unknown": "Unknown",
     "Unload model to save GPU memory": "卸載音色節省 VRAM",
     "Version": "版本",

i18n/locale/zh_SG.json

@@ -140,7 +140,6 @@
     "Train model": "訓練模型",
     "Training complete. You can check the training logs in the console or the 'train.log' file under the experiment folder.": "训练结束, 您可查看控制台训练日志或实验文件夹下的train.log",
     "Transpose (integer, number of semitones, raise by an octave: 12, lower by an octave: -12)": "變調(整數、半音數量、升八度12降八度-12)",
-    "Unfortunately, there is no compatible GPU available to support your training.": "很遗憾您这没有能用的显卡来支持您训练",
     "Unknown": "Unknown",
     "Unload model to save GPU memory": "卸載音色節省 VRAM",
     "Version": "版本",

i18n/locale/zh_TW.json

@@ -140,7 +140,6 @@
     "Train model": "訓練模型",
     "Training complete. You can check the training logs in the console or the 'train.log' file under the experiment folder.": "训练结束, 您可查看控制台训练日志或实验文件夹下的train.log",
     "Transpose (integer, number of semitones, raise by an octave: 12, lower by an octave: -12)": "變調(整數、半音數量、升八度12降八度-12)",
-    "Unfortunately, there is no compatible GPU available to support your training.": "很遗憾您这没有能用的显卡来支持您训练",
     "Unknown": "Unknown",
     "Unload model to save GPU memory": "卸載音色節省 VRAM",
     "Version": "版本",

infer/lib/audio.py

@@ -195,7 +195,10 @@ def get_audio_properties(input_path: str) -> Tuple[int, int]:
     container = av.open(input_path)
     audio_stream = next(s for s in container.streams if s.type == "audio")
     channels = 1 if audio_stream.layout == "mono" else 2
-    rate = audio_stream.base_rate
+    try:
+        rate = audio_stream.base_rate
+    except:
+        rate = audio_stream.sample_rate
     container.close()
     return channels, rate
 

infer/lib/rvcmd.py

@@ -162,15 +162,7 @@ def download_and_extract_zip(url: str, folder: str):
         logger.info(f"extracted into {folder}")
 
 
-def download_dns_yaml(url: str, folder: str):
-    logger.info(f"downloading {url}")
-    response = requests.get(url, stream=True, timeout=(5, 10))
-    with open(os.path.join(folder, "dns.yaml"), "wb") as out_file:
-        out_file.write(response.content)
-        logger.info(f"downloaded into {folder}")
-
-
-def download_all_assets(tmpdir: str, version="0.2.5"):
+def download_all_assets(tmpdir: str, version="0.2.11"):
     import subprocess
     import platform
 
@@ -198,44 +190,10 @@ def download_all_assets(tmpdir: str, version="0.2.5"):
     suffix = "zip" if is_win else "tar.gz"
     RVCMD_URL = BASE_URL + f"v{version}/rvcmd_{system_type}_{architecture}.{suffix}"
     cmdfile = os.path.join(tmpdir, "rvcmd")
-    try:
-        if is_win:
-            download_and_extract_zip(RVCMD_URL, tmpdir)
-            cmdfile += ".exe"
-        else:
-            download_and_extract_tar_gz(RVCMD_URL, tmpdir)
-            os.chmod(cmdfile, 0o755)
-        subprocess.run([cmdfile, "-notui", "-w", "0", "assets/rvc"])
-    except Exception:
-        BASE_URL = "https://raw.gitcode.com/u011570312/RVC-Models-Downloader/assets/"
-        suffix = {
-            "darwin_amd64": "555",
-            "darwin_arm64": "556",
-            "linux_386": "557",
-            "linux_amd64": "558",
-            "linux_arm64": "559",
-            "windows_386": "562",
-            "windows_amd64": "563",
-        }[f"{system_type}_{architecture}"]
-        RVCMD_URL = BASE_URL + suffix
-        download_dns_yaml(
-            "https://raw.gitcode.com/u011570312/RVC-Models-Downloader/raw/main/dns.yaml",
-            tmpdir,
-        )
-        if is_win:
-            download_and_extract_zip(RVCMD_URL, tmpdir)
-            cmdfile += ".exe"
-        else:
-            download_and_extract_tar_gz(RVCMD_URL, tmpdir)
-            os.chmod(cmdfile, 0o755)
-        subprocess.run(
-            [
-                cmdfile,
-                "-notui",
-                "-w",
-                "0",
-                "-dns",
-                os.path.join(tmpdir, "dns.yaml"),
-                "assets/rvc",
-            ]
-        )
+    if is_win:
+        download_and_extract_zip(RVCMD_URL, tmpdir)
+        cmdfile += ".exe"
+    else:
+        download_and_extract_tar_gz(RVCMD_URL, tmpdir)
+        os.chmod(cmdfile, 0o755)
+    subprocess.run([cmdfile, "-notui", "-w", "0", "assets/rvc"])

infer/lib/train/data_utils.py

@@ -303,7 +303,7 @@ class TextAudioLoader(torch.utils.data.Dataset):
         spec_filename = filename.replace(".wav", ".spec.pt")
         if os.path.exists(spec_filename):
             try:
-                spec = torch.load(spec_filename)
+                spec = torch.load(spec_filename, weights_only=True)
             except:
                 logger.warning("%s %s", spec_filename, traceback.format_exc())
                 spec = spectrogram_torch(

infer/lib/train/process_ckpt.py

@@ -59,7 +59,7 @@ def save_small_model(ckpt, sr, if_f0, name, epoch, version, hps):
 
 def extract_small_model(path, name, author, sr, if_f0, info, version):
     try:
-        ckpt = torch.load(path, map_location="cpu")
+        ckpt = torch.load(path, map_location="cpu", weights_only=True)
         if "model" in ckpt:
             ckpt = ckpt["model"]
         opt = OrderedDict()
@@ -196,7 +196,7 @@ def extract_small_model(path, name, author, sr, if_f0, info, version):
 
 def change_info(path, info, name):
     try:
-        ckpt = torch.load(path, map_location="cpu")
+        ckpt = torch.load(path, map_location="cpu", weights_only=True)
         ckpt["info"] = info
         if name == "":
             name = os.path.basename(path)
@@ -229,8 +229,8 @@ def merge(path1, path2, alpha1, sr, f0, info, name, version):
                 a2 = "Unknown"
             return f"{a1} & {a2}"
 
-        ckpt1 = torch.load(path1, map_location="cpu")
-        ckpt2 = torch.load(path2, map_location="cpu")
+        ckpt1 = torch.load(path1, map_location="cpu", weights_only=True)
+        ckpt2 = torch.load(path2, map_location="cpu", weights_only=True)
         cfg = ckpt1["config"]
         if "model" in ckpt1:
             ckpt1 = extract(ckpt1)

infer/lib/train/utils.py

@@ -29,6 +29,24 @@ def load_checkpoint(checkpoint_path, model, optimizer=None, load_opt=1):
     checkpoint_dict = torch.load(checkpoint_path, map_location="cpu", weights_only=True)
 
     saved_state_dict = checkpoint_dict["model"]
+    # Convert old-style weight_norm keys (weight_g/weight_v) to new
+    # parametrizations format (parametrizations.weight.original0/original1)
+    # so that checkpoints saved with the deprecated API can still be loaded.
+    _converted = {}
+    for k, v in list(saved_state_dict.items()):
+        if k.endswith(".weight_g"):
+            new_key = k[: -len(".weight_g")] + ".parametrizations.weight.original0"
+            _converted[new_key] = v
+        elif k.endswith(".weight_v"):
+            new_key = k[: -len(".weight_v")] + ".parametrizations.weight.original1"
+            _converted[new_key] = v
+    if _converted:
+        logger.info(
+            "Converting %d old-style weight_norm keys from checkpoint to new parametrizations format",
+            len(_converted),
+        )
+        saved_state_dict.update(_converted)
+
     if hasattr(model, "module"):
         state_dict = model.module.state_dict()
     else:

infer/modules/train/extract_f0_print.py

@@ -2,6 +2,7 @@ import os
 import sys
 import traceback
 from pathlib import Path
+import importlib.util
 
 from dotenv import load_dotenv
 
@@ -38,6 +39,9 @@ f0method = sys.argv[3]
 device = sys.argv[4]
 is_half = sys.argv[5] == "True"
 
+if importlib.util.find_spec("torch_directml") is not None:
+    import torch_directml  # use side effect
+
 
 class FeatureInput(object):
     def __init__(self, is_half: bool, device="cpu", samplerate=16000, hop_size=160):
@@ -102,6 +106,12 @@ if __name__ == "__main__":
     Config.use_insecure_load()
 
     printt(" ".join(sys.argv))
+    # GPU methods (rmvpe, fcpe, crepe, etc.) gain nothing from multiprocessing since
+    # all processes share one GPU. Spawning n_p processes each lazily loading
+    # the model onto the same CUDA device exhausts VRAM and causes deadlocks.
+    if "cuda" in device:
+        printt("WARN: use 1 thread since GPU is used.")
+        n_p = 1
     featureInput = FeatureInput(is_half, device)
     paths = []
     inp_root = "%s/1_16k_wavs" % (exp_dir)

infer/modules/train/extract_feature_print.py

@@ -17,7 +17,14 @@ device = sys.argv[1]
 n_part = int(sys.argv[2])
 i_part = int(sys.argv[3])
 i_gpu = sys.argv[4]
-os.environ["CUDA_VISIBLE_DEVICES"] = str(i_gpu)
+# CUDA_VISIBLE_DEVICES expects bare GPU indices (e.g. "0" or "0,1"),
+# but callers may pass "cuda:0", "cuda:0-cuda:1", etc.  Strip the prefix
+# and normalise separators so any combination works.
+import re
+
+i_gpu = re.sub(r"cuda:", "", str(i_gpu))
+i_gpu = i_gpu.replace("-", ",")
+os.environ["CUDA_VISIBLE_DEVICES"] = i_gpu
 exp_dir = sys.argv[5]
 version = sys.argv[6]
 is_half = sys.argv[7].lower() == "true"

infer/modules/train/train.py

@@ -29,10 +29,12 @@ try:
 
         GradScaler = gradscaler_init()
         ipex_init()
-    else:
-        from torch.cuda.amp import GradScaler, autocast
 except Exception:
-    from torch.cuda.amp import GradScaler, autocast
+    pass
+finally:
+    if not ("GradScaler" in globals() and "autocast" in globals()):
+        from torch.amp.grad_scaler import GradScaler
+        from torch.amp.autocast_mode import autocast
 
 torch.backends.cudnn.deterministic = False
 torch.backends.cudnn.benchmark = False
@@ -106,23 +108,30 @@ def main():
         # patch to unblock people without gpus. there is probably a better way.
         print("NO GPU DETECTED: falling back to CPU - this may take a while")
         n_gpus = 1
-    os.environ["MASTER_ADDR"] = "localhost"
-    os.environ["MASTER_PORT"] = str(randint(20000, 55555))
-    children = []
     logger = utils.get_logger(hps.model_dir)
-    for i in range(n_gpus):
-        subproc = mp.Process(
-            target=run,
-            args=(i, n_gpus, hps, logger),
-        )
-        children.append(subproc)
-        subproc.start()
+    if n_gpus == 1:
+        # Single GPU: run directly without distributed to avoid gloo issues on Windows
+        run(0, 1, hps, logger)
+    else:
+        master_port = str(randint(20000, 55555))
+        os.environ["MASTER_ADDR"] = "127.0.0.1"
+        os.environ["MASTER_PORT"] = master_port
+        children = []
+        for i in range(n_gpus):
+            subproc = mp.Process(
+                target=run,
+                args=(i, n_gpus, hps, logger, master_port),
+            )
+            children.append(subproc)
+            subproc.start()
 
-    for i in range(n_gpus):
-        children[i].join()
+        for i in range(n_gpus):
+            children[i].join()
 
 
-def run(rank, n_gpus, hps: utils.HParams, logger: logging.Logger):
+def run(
+    rank, n_gpus, hps: utils.HParams, logger: logging.Logger, master_port: str = "29500"
+):
     global global_step
     if rank == 0:
         # logger = utils.get_logger(hps.model_dir)
@@ -131,24 +140,81 @@ def run(rank, n_gpus, hps: utils.HParams, logger: logging.Logger):
         writer = SummaryWriter(log_dir=hps.model_dir)
         writer_eval = SummaryWriter(log_dir=os.path.join(hps.model_dir, "eval"))
 
-    try:
-        dist.init_process_group(
-            backend=(
-                "gloo" if os.name == "nt" or not torch.cuda.is_available() else "nccl"
-            ),
-            init_method="env://",
-            world_size=n_gpus,
-            rank=rank,
-        )
-    except:
-        dist.init_process_group(
-            backend=(
-                "gloo" if os.name == "nt" or not torch.cuda.is_available() else "nccl"
-            ),
-            init_method="env://?use_libuv=False",
-            world_size=n_gpus,
-            rank=rank,
-        )
+    use_distributed = n_gpus > 1
+    if use_distributed:
+        if os.name == "nt" or not torch.cuda.is_available():
+            # On Windows, gloo's create_device(hostname=...) is gated to Linux only
+            # in the C++ layer (makeDeviceForHostname). We must use the interface-
+            # based path instead: create_device(interface=...) calls
+            # makeDeviceForInterface which is not platform-gated.
+            import socket as _socket
+
+            try:
+                store = dist.TCPStore(
+                    host_name="127.0.0.1",
+                    port=int(master_port),
+                    world_size=n_gpus,
+                    is_master=(rank == 0),
+                )
+            except Exception:
+                store = dist.TCPStore(
+                    host_name="127.0.0.1",
+                    port=int(master_port),
+                    world_size=n_gpus,
+                    is_master=(rank == 0),
+                    use_libuv=False,
+                )
+
+            # Discover a working network interface for gloo device creation
+            gloo_device = None
+            try:
+                for idx, ifname in _socket.if_nameindex():
+                    try:
+                        gloo_device = dist.ProcessGroupGloo.create_device(
+                            interface=ifname
+                        )
+                        print("Try device", idx, "name", ifname)
+                        break
+                    except RuntimeError as e:
+                        print("Try device", idx, "name", ifname, "err:", e)
+                        continue
+            except (OSError, AttributeError) as e:
+                print(e.with_traceback(None))
+
+            if gloo_device is None:
+                raise RuntimeError(
+                    "Cannot create gloo device on Windows. "
+                    "No usable network interface found. "
+                    "Try adding your hostname to "
+                    "C:\\Windows\\System32\\drivers\\etc\\hosts "
+                    "with: 127.0.0.1  " + _socket.gethostname()
+                )
+
+            pg_options = dist.ProcessGroupGloo._Options()
+            pg_options._devices = [gloo_device]
+            dist.init_process_group(
+                backend="gloo",
+                store=store,
+                world_size=n_gpus,
+                rank=rank,
+                pg_options=pg_options,
+            )
+        else:
+            init_url = f"tcp://127.0.0.1:{master_port}"
+            try:
+                dist.init_process_group(
+                    backend="nccl",
+                    init_method=init_url,
+                    world_size=n_gpus,
+                    rank=rank,
+                )
+            except:
+                dist.init_process_group(
+                    backend="nccl",
+                    init_method=init_url + "?use_libuv=False",
+                    world_size=n_gpus,
+                    rank=rank,
+                )
     torch.manual_seed(hps.train.seed)
     if torch.cuda.is_available():
         torch.cuda.set_device(rank)
@@ -221,14 +287,15 @@ def run(rank, n_gpus, hps: utils.HParams, logger: logging.Logger):
     )
     # net_g = DDP(net_g, device_ids=[rank], find_unused_parameters=True)
     # net_d = DDP(net_d, device_ids=[rank], find_unused_parameters=True)
-    if hasattr(torch, "xpu") and torch.xpu.is_available():
-        pass
-    elif torch.cuda.is_available():
-        net_g = DDP(net_g, device_ids=[rank])
-        net_d = DDP(net_d, device_ids=[rank])
-    else:
-        net_g = DDP(net_g)
-        net_d = DDP(net_d)
+    if use_distributed:
+        if hasattr(torch, "xpu") and torch.xpu.is_available():
+            pass
+        elif torch.cuda.is_available():
+            net_g = DDP(net_g, device_ids=[rank])
+            net_d = DDP(net_d, device_ids=[rank])
+        else:
+            net_g = DDP(net_g)
+            net_d = DDP(net_d)
 
     try:  # 如果能加载自动resume
         _, _, _, epoch_str = utils.load_checkpoint(
@@ -470,7 +537,7 @@ def train_and_evaluate(
             # wave_lengths = wave_lengths.cuda(rank, non_blocking=True)
 
         # Calculate
-        with autocast(enabled=hps.train.fp16_run):
+        with autocast(device_type="cuda", enabled=hps.train.fp16_run):
             (
                 y_hat,
                 ids_slice,
@@ -489,7 +556,7 @@ def train_and_evaluate(
             y_mel = slice_on_last_dim(
                 mel, ids_slice, hps.train.segment_size // hps.data.hop_length
             )
-            with autocast(enabled=False):
+            with autocast(device_type="cuda", enabled=False):
                 y_hat_mel = mel_spectrogram_torch(
                     y_hat.float().squeeze(1),
                     hps.data.filter_length,
@@ -508,7 +575,7 @@ def train_and_evaluate(
 
             # Discriminator
             y_d_hat_r, y_d_hat_g, _, _ = net_d(wave, y_hat.detach())
-            with autocast(enabled=False):
+            with autocast(device_type="cuda", enabled=False):
                 loss_disc, losses_disc_r, losses_disc_g = discriminator_loss(
                     y_d_hat_r, y_d_hat_g
                 )
@@ -518,10 +585,10 @@ def train_and_evaluate(
         grad_norm_d = total_grad_norm(net_d.parameters())
         scaler.step(optim_d)
 
-        with autocast(enabled=hps.train.fp16_run):
+        with autocast(device_type="cuda", enabled=hps.train.fp16_run):
             # Generator
             y_d_hat_r, y_d_hat_g, fmap_r, fmap_g = net_d(wave, y_hat)
-            with autocast(enabled=False):
+            with autocast(device_type="cuda", enabled=False):
                 loss_mel = F.l1_loss(y_mel, y_hat_mel) * hps.train.c_mel
                 loss_kl = kl_loss(z_p, logs_q, m_p, logs_p, z_mask) * hps.train.c_kl
                 loss_fm = feature_loss(fmap_r, fmap_g)

infer/modules/uvr5/modules.py

@@ -62,10 +62,6 @@ def uvr(model_name, inp_root, save_root_vocal, paths, save_root_ins, agg, format
                     os.path.basename(inp_path),
                 )
                 resample_audio(inp_path, tmp_path, "pcm_s16le", "s16", 44100, "stereo")
-                try:  # Remove the original file
-                    os.remove(inp_path)
-                except Exception as e:
-                    print(f"Failed to remove the original file: {e}")
                 inp_path = tmp_path
             try:
                 if done == 0:

infer/modules/uvr5/vr.py

@@ -37,7 +37,7 @@ class AudioPre:
         else:
             mp = ModelParameters("infer/lib/uvr5_pack/lib_v5/modelparams/4band_v2.json")
             model = Nets.CascadedASPPNet(mp.param["bins"] * 2)
-        cpk = torch.load(model_path, map_location="cpu")
+        cpk = torch.load(model_path, map_location="cpu", weights_only=True)
         model.load_state_dict(cpk)
         model.eval()
         if is_half:

infer/modules/vc/hash.py

@@ -10,9 +10,6 @@ from pybase16384 import encode_to_string, decode_from_string
 from configs import CPUConfig
 from rvc.synthesizer import get_synthesizer
 
-from .pipeline import Pipeline
-from .utils import load_hubert
-
 
 class TorchSeedContext:
     def __init__(self, seed):
@@ -95,6 +92,9 @@ def wave_hash(time_field):
 
 
 def model_hash(config, tgt_sr, net_g, if_f0, version):
+    from .pipeline import Pipeline
+    from .utils import load_hubert
+
     pipeline = Pipeline(tgt_sr, config)
     audio = original_audio()
     hbt = load_hubert(config.device, config.is_half)
@@ -152,7 +152,7 @@ def model_hash_ckpt(cpt):
 
 
 def model_hash_from(path):
-    cpt = torch.load(path, map_location="cpu")
+    cpt = torch.load(path, map_location="cpu", weights_only=True)
     h = model_hash_ckpt(cpt)
     del cpt
     return h

infer/modules/vc/info.py

@@ -75,7 +75,7 @@ def show_info(path):
     try:
         if hasattr(path, "name"):
             path = path.name
-        a = torch.load(path, map_location="cpu")
+        a = torch.load(path, map_location="cpu", weights_only=True)
         txt = show_model_info(a, show_long_id=True)
         del a
     except:

infer/modules/vc/utils.py

@@ -1,6 +1,7 @@
 import os, pathlib
 
-from fairseq import checkpoint_utils
+import torch
+from fairseq import checkpoint_utils, data
 
 
 def get_index_path_from_model(sid):
@@ -21,10 +22,11 @@ def get_index_path_from_model(sid):
 
 
 def load_hubert(device, is_half):
-    models, _, _ = checkpoint_utils.load_model_ensemble_and_task(
-        ["assets/hubert/hubert_base.pt"],
-        suffix="",
-    )
+    with torch.serialization.safe_globals([data.dictionary.Dictionary]):
+        models, _, _ = checkpoint_utils.load_model_ensemble_and_task(
+            ["assets/hubert/hubert_base.pt"],
+            suffix="",
+        )
     hubert_model = models[0]
     hubert_model = hubert_model.to(device)
     if is_half:

requirements/amd.txt

@@ -1,11 +1,10 @@
-tensorflow-rocm
 joblib>=1.1.0
 numba
 numpy
 scipy
 librosa>=0.10.2
 llvmlite
-fairseq @ git+https://github.com/One-sixth/fairseq.git
+fairseq @ git+https://github.com/fumiama/fairseq.git
 faiss-cpu
 gradio
 Cython

requirements/cpu.txt

@@ -0,0 +1,49 @@
+joblib>=1.1.0
+numba
+numpy
+scipy
+librosa>=0.10.2
+llvmlite
+fairseq @ git+https://github.com/fumiama/fairseq.git
+faiss-cpu
+gradio
+Cython
+pydub>=0.25.1
+tensorboardX
+Jinja2>=3.1.2
+json5
+Markdown
+matplotlib>=3.7.0
+matplotlib-inline>=0.1.3
+praat-parselmouth>=0.4.2
+Pillow>=9.1.1
+resampy>=0.4.2
+scikit-learn
+tensorboard
+tqdm>=4.63.1
+tornado>=6.1
+Werkzeug>=2.2.3
+uc-micro-py>=1.0.1
+sympy>=1.11.1
+tabulate>=0.8.10
+PyYAML>=6.0
+pyasn1>=0.4.8
+pyasn1-modules>=0.2.8
+fsspec>=2022.11.0
+absl-py>=1.2.0
+audioread
+uvicorn>=0.21.1
+colorama>=0.4.5
+pyworld==0.3.2
+httpx
+onnxruntime; sys_platform == 'darwin'
+torchcrepe>=0.0.23
+fastapi
+torchfcpe
+python-dotenv>=1.0.0
+av
+pybase16384
+--extra-index-url https://download.pytorch.org/whl/cpu
+torch
+torchvision
+torchaudio

requirements/dml.txt

@@ -4,7 +4,7 @@ numpy
 scipy
 librosa>=0.10.2
 llvmlite
-fairseq @ git+https://github.com/One-sixth/fairseq.git
+fairseq @ git+https://github.com/fumiama/fairseq.git
 faiss-cpu
 gradio
 Cython
@@ -43,3 +43,6 @@ python-dotenv>=1.0.0
 av
 torchfcpe
 pybase16384
+torch-directml
+torchvision
+torchaudio

requirements/ipex.txt

@@ -9,7 +9,7 @@ numpy
 scipy
 librosa>=0.10.2
 llvmlite==0.39.0
-fairseq @ git+https://github.com/One-sixth/fairseq.git
+fairseq @ git+https://github.com/fumiama/fairseq.git
 faiss-cpu
 gradio
 Cython

requirements/main.txt

@@ -4,7 +4,7 @@ numpy
 scipy
 librosa>=0.10.2
 llvmlite
-fairseq @ git+https://github.com/One-sixth/fairseq.git
+fairseq @ git+https://github.com/fumiama/fairseq.git
 faiss-cpu
 gradio
 Cython

rvc/f0/f0.py

@@ -11,14 +11,14 @@ class F0Predictor(object):
         f0_min=50,
         f0_max=1100,
         sampling_rate=44100,
-        device: Optional[str] = None,
+        device: Optional[Union[str, torch.device]] = None,
     ):
         self.hop_length = hop_length
         self.f0_min = f0_min
         self.f0_max = f0_max
         self.sampling_rate = sampling_rate
-        if device is None:
-            device = "cuda:0" if torch.cuda.is_available() else "cpu"
+        if not device:
+            device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
         self.device = device
 
     def compute_f0(

rvc/f0/mel.py

@@ -1,4 +1,4 @@
-from typing import Optional
+from typing import Optional, Union
 
 import torch
 import numpy as np
@@ -17,9 +17,9 @@ class MelSpectrogram(torch.nn.Module):
         hop_length: int,
         n_fft: Optional[int] = None,
         mel_fmin: int = 0,
-        mel_fmax: int = None,
+        mel_fmax: Optional[int] = None,
         clamp: float = 1e-5,
-        device=torch.device("cpu"),
+        device: Union[str, torch.device] = torch.device("cpu"),
     ):
         super().__init__()
         if n_fft is None:

rvc/f0/rmvpe.py

@@ -1,6 +1,6 @@
 from io import BytesIO
 import os
-from typing import Any, Optional, Union
+from typing import Optional, Union
 
 import numpy as np
 import torch

rvc/layers/attentions.py

@@ -12,8 +12,8 @@ class MultiHeadAttention(nn.Module):
         channels: int,
         out_channels: int,
         n_heads: int,
+        window_size: int,
         p_dropout: float = 0.0,
-        window_size: Optional[int] = None,
         heads_share: bool = True,
         block_length: Optional[int] = None,
         proximal_bias: bool = False,

rvc/layers/discriminators.py

@@ -4,7 +4,8 @@ import torch
 from torch import nn
 from torch.nn import Conv1d, Conv2d
 from torch.nn import functional as F
-from torch.nn.utils import spectral_norm, weight_norm
+from torch.nn.utils import spectral_norm
+from torch.nn.utils.parametrizations import weight_norm
 
 from .residuals import LRELU_SLOPE
 from .utils import get_padding

rvc/layers/encoders.py

@@ -42,8 +42,8 @@ class Encoder(nn.Module):
                     hidden_channels,
                     hidden_channels,
                     n_heads,
+                    window_size,
                     p_dropout=p_dropout,
-                    window_size=window_size,
                 )
             )
             self.norm_layers_1.append(LayerNorm(hidden_channels))
@@ -121,7 +121,7 @@ class TextEncoder(nn.Module):
     def __call__(
         self,
         phone: torch.Tensor,
-        pitch: torch.Tensor,
+        pitch: Optional[torch.Tensor],
         lengths: torch.Tensor,
         skip_head: Optional[int] = None,
     ) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
@@ -135,7 +135,7 @@ class TextEncoder(nn.Module):
     def forward(
         self,
         phone: torch.Tensor,
-        pitch: torch.Tensor,
+        pitch: Optional[torch.Tensor],
         lengths: torch.Tensor,
         skip_head: Optional[int] = None,
     ) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
@@ -212,10 +212,8 @@ class PosteriorEncoder(nn.Module):
         self.enc.remove_weight_norm()
 
     def __prepare_scriptable__(self):
-        for hook in self.enc._forward_pre_hooks.values():
-            if (
-                hook.__module__ == "torch.nn.utils.weight_norm"
-                and hook.__class__.__name__ == "WeightNorm"
-            ):
-                torch.nn.utils.remove_weight_norm(self.enc)
+        from torch.nn.utils import parametrize
+
+        if parametrize.is_parametrized(self.enc, "weight"):
+            parametrize.remove_parametrizations(self.enc, "weight")
         return self

rvc/layers/generators.py

@@ -4,7 +4,8 @@ import torch
 from torch import nn
 from torch.nn import Conv1d, ConvTranspose1d
 from torch.nn import functional as F
-from torch.nn.utils import remove_weight_norm, weight_norm
+from torch.nn.utils.parametrizations import weight_norm
+from torch.nn.utils.parametrize import is_parametrized, remove_parametrizations
 
 from .residuals import ResBlock1, ResBlock2, LRELU_SLOPE
 from .utils import call_weight_data_normal_if_Conv
@@ -46,6 +47,7 @@ class Generator(torch.nn.Module):
 
         self.resblocks = nn.ModuleList()
         resblock_module = ResBlock1 if resblock == "1" else ResBlock2
+        ch = 0
         for i in range(len(self.ups)):
             ch = upsample_initial_channel // (2 ** (i + 1))
             for k, d in zip(resblock_kernel_sizes, resblock_dilation_sizes):
@@ -97,29 +99,16 @@ class Generator(torch.nn.Module):
 
     def __prepare_scriptable__(self):
         for l in self.ups:
-            for hook in l._forward_pre_hooks.values():
-                # The hook we want to remove is an instance of WeightNorm class, so
-                # normally we would do `if isinstance(...)` but this class is not accessible
-                # because of shadowing, so we check the module name directly.
-                # https://github.com/pytorch/pytorch/blob/be0ca00c5ce260eb5bcec3237357f7a30cc08983/torch/nn/utils/__init__.py#L3
-                if (
-                    hook.__module__ == "torch.nn.utils.weight_norm"
-                    and hook.__class__.__name__ == "WeightNorm"
-                ):
-                    torch.nn.utils.remove_weight_norm(l)
-
+            if is_parametrized(l, "weight"):
+                remove_parametrizations(l, "weight")
         for l in self.resblocks:
-            for hook in l._forward_pre_hooks.values():
-                if (
-                    hook.__module__ == "torch.nn.utils.weight_norm"
-                    and hook.__class__.__name__ == "WeightNorm"
-                ):
-                    torch.nn.utils.remove_weight_norm(l)
+            if is_parametrized(l, "weight"):
+                remove_parametrizations(l, "weight")
         return self
 
     def remove_weight_norm(self):
         for l in self.ups:
-            remove_weight_norm(l)
+            remove_parametrizations(l, "weight")
         for l in self.resblocks:
             l.remove_weight_norm()
 

rvc/layers/norms.py

@@ -6,6 +6,8 @@ from torch.nn import functional as F
 
 from .utils import activate_add_tanh_sigmoid_multiply
 
+from torch.nn.utils.parametrize import is_parametrized, remove_parametrizations
+
 
 class LayerNorm(nn.Module):
     def __init__(self, channels: int, eps: float = 1e-5):
@@ -30,7 +32,7 @@ class WN(torch.nn.Module):
         dilation_rate: int,
         n_layers: int,
         gin_channels: int = 0,
-        p_dropout: int = 0,
+        p_dropout: float = 0,
     ):
         super(WN, self).__init__()
         assert kernel_size % 2 == 1
@@ -49,7 +51,9 @@ class WN(torch.nn.Module):
             cond_layer = torch.nn.Conv1d(
                 gin_channels, 2 * hidden_channels * n_layers, 1
             )
-            self.cond_layer = torch.nn.utils.weight_norm(cond_layer, name="weight")
+            self.cond_layer = torch.nn.utils.parametrizations.weight_norm(
+                cond_layer, name="weight"
+            )
 
         for i in range(n_layers):
             dilation = dilation_rate**i
@@ -61,7 +65,9 @@ class WN(torch.nn.Module):
                 dilation=dilation,
                 padding=padding,
             )
-            in_layer = torch.nn.utils.weight_norm(in_layer, name="weight")
+            in_layer = torch.nn.utils.parametrizations.weight_norm(
+                in_layer, name="weight"
+            )
             self.in_layers.append(in_layer)
 
             # last one is not necessary
@@ -71,7 +77,9 @@ class WN(torch.nn.Module):
                 res_skip_channels = hidden_channels
 
             res_skip_layer = torch.nn.Conv1d(hidden_channels, res_skip_channels, 1)
-            res_skip_layer = torch.nn.utils.weight_norm(res_skip_layer, name="weight")
+            res_skip_layer = torch.nn.utils.parametrizations.weight_norm(
+                res_skip_layer, name="weight"
+            )
             self.res_skip_layers.append(res_skip_layer)
 
     def __call__(
@@ -117,32 +125,20 @@ class WN(torch.nn.Module):
 
     def remove_weight_norm(self):
         if self.gin_channels != 0:
-            torch.nn.utils.remove_weight_norm(self.cond_layer)
+            remove_parametrizations(self.cond_layer, "weight")
         for l in self.in_layers:
-            torch.nn.utils.remove_weight_norm(l)
+            remove_parametrizations(l, "weight")
         for l in self.res_skip_layers:
-            torch.nn.utils.remove_weight_norm(l)
+            remove_parametrizations(l, "weight")
 
     def __prepare_scriptable__(self):
         if self.gin_channels != 0:
-            for hook in self.cond_layer._forward_pre_hooks.values():
-                if (
-                    hook.__module__ == "torch.nn.utils.weight_norm"
-                    and hook.__class__.__name__ == "WeightNorm"
-                ):
-                    torch.nn.utils.remove_weight_norm(self.cond_layer)
+            if is_parametrized(self.cond_layer, "weight"):
+                remove_parametrizations(self.cond_layer, "weight")
         for l in self.in_layers:
-            for hook in l._forward_pre_hooks.values():
-                if (
-                    hook.__module__ == "torch.nn.utils.weight_norm"
-                    and hook.__class__.__name__ == "WeightNorm"
-                ):
-                    torch.nn.utils.remove_weight_norm(l)
+            if is_parametrized(l, "weight"):
+                remove_parametrizations(l, "weight")
         for l in self.res_skip_layers:
-            for hook in l._forward_pre_hooks.values():
-                if (
-                    hook.__module__ == "torch.nn.utils.weight_norm"
-                    and hook.__class__.__name__ == "WeightNorm"
-                ):
-                    torch.nn.utils.remove_weight_norm(l)
+            if is_parametrized(l, "weight"):
+                remove_parametrizations(l, "weight")
         return self

rvc/layers/nsf.py

@@ -1,11 +1,12 @@
-from typing import Optional, List
+from typing import Optional, List, Union
 import math
 
 import torch
 from torch import nn
 from torch.nn import Conv1d, ConvTranspose1d
 from torch.nn import functional as F
-from torch.nn.utils import remove_weight_norm, weight_norm
+from torch.nn.utils.parametrizations import weight_norm
+from torch.nn.utils.parametrize import is_parametrized, remove_parametrizations
 
 from .generators import SineGenerator
 from .residuals import ResBlock1, ResBlock2, LRELU_SLOPE
@@ -83,7 +84,7 @@ class NSFGenerator(torch.nn.Module):
         self.conv_pre = Conv1d(
             initial_channel, upsample_initial_channel, 7, 1, padding=3
         )
-        resblock = ResBlock1 if resblock == "1" else ResBlock2
+        resblockcls = ResBlock1 if resblock == "1" else ResBlock2
 
         self.ups = nn.ModuleList()
         for i, (u, k) in enumerate(zip(upsample_rates, upsample_kernel_sizes)):
@@ -114,12 +115,13 @@ class NSFGenerator(torch.nn.Module):
                 self.noise_convs.append(Conv1d(1, c_cur, kernel_size=1))
 
         self.resblocks = nn.ModuleList()
+        ch = 0
         for i in range(len(self.ups)):
-            ch: int = upsample_initial_channel // (2 ** (i + 1))
+            ch = upsample_initial_channel // (2 ** (i + 1))
             for j, (k, d) in enumerate(
                 zip(resblock_kernel_sizes, resblock_dilation_sizes)
             ):
-                self.resblocks.append(resblock(ch, k, d))
+                self.resblocks.append(resblockcls(ch, k, d))
 
         self.conv_post = Conv1d(ch, 1, 7, 1, padding=3, bias=False)
         self.ups.apply(call_weight_data_normal_if_Conv)
@@ -190,27 +192,15 @@ class NSFGenerator(torch.nn.Module):
 
     def remove_weight_norm(self):
         for l in self.ups:
-            remove_weight_norm(l)
+            remove_parametrizations(l, "weight")
         for l in self.resblocks:
             l.remove_weight_norm()
 
     def __prepare_scriptable__(self):
         for l in self.ups:
-            for hook in l._forward_pre_hooks.values():
-                # The hook we want to remove is an instance of WeightNorm class, so
-                # normally we would do `if isinstance(...)` but this class is not accessible
-                # because of shadowing, so we check the module name directly.
-                # https://github.com/pytorch/pytorch/blob/be0ca00c5ce260eb5bcec3237357f7a30cc08983/torch/nn/utils/__init__.py#L3
-                if (
-                    hook.__module__ == "torch.nn.utils.weight_norm"
-                    and hook.__class__.__name__ == "WeightNorm"
-                ):
-                    torch.nn.utils.remove_weight_norm(l)
+            if is_parametrized(l, "weight"):
+                remove_parametrizations(l, "weight")
         for l in self.resblocks:
-            for hook in self.resblocks._forward_pre_hooks.values():
-                if (
-                    hook.__module__ == "torch.nn.utils.weight_norm"
-                    and hook.__class__.__name__ == "WeightNorm"
-                ):
-                    torch.nn.utils.remove_weight_norm(l)
+            if is_parametrized(l, "weight"):
+                remove_parametrizations(l, "weight")
         return self

rvc/layers/residuals.py

@@ -4,7 +4,8 @@ import torch
 from torch import nn
 from torch.nn import Conv1d
 from torch.nn import functional as F
-from torch.nn.utils import remove_weight_norm, weight_norm
+from torch.nn.utils.parametrizations import weight_norm
+from torch.nn.utils.parametrize import is_parametrized, remove_parametrizations
 
 from .norms import WN
 from .utils import (
@@ -20,7 +21,7 @@ class ResBlock1(torch.nn.Module):
         self,
         channels: int,
         kernel_size: int = 3,
-        dilation: List[int] = (1, 3, 5),
+        dilation: List[int] = [1, 3, 5],
     ):
         super(ResBlock1, self).__init__()
 
@@ -85,25 +86,17 @@ class ResBlock1(torch.nn.Module):
 
     def remove_weight_norm(self):
         for l in self.convs1:
-            remove_weight_norm(l)
+            remove_parametrizations(l, "weight")
         for l in self.convs2:
-            remove_weight_norm(l)
+            remove_parametrizations(l, "weight")
 
     def __prepare_scriptable__(self):
         for l in self.convs1:
-            for hook in l._forward_pre_hooks.values():
-                if (
-                    hook.__module__ == "torch.nn.utils.weight_norm"
-                    and hook.__class__.__name__ == "WeightNorm"
-                ):
-                    torch.nn.utils.remove_weight_norm(l)
+            if is_parametrized(l, "weight"):
+                remove_parametrizations(l, "weight")
         for l in self.convs2:
-            for hook in l._forward_pre_hooks.values():
-                if (
-                    hook.__module__ == "torch.nn.utils.weight_norm"
-                    and hook.__class__.__name__ == "WeightNorm"
-                ):
-                    torch.nn.utils.remove_weight_norm(l)
+            if is_parametrized(l, "weight"):
+                remove_parametrizations(l, "weight")
         return self
 
 
@@ -117,7 +110,7 @@ class ResBlock2(torch.nn.Module):
         self,
         channels: int,
         kernel_size=3,
-        dilation: List[int] = (1, 3),
+        dilation: List[int] = [1, 3],
     ):
         super(ResBlock2, self).__init__()
         self.convs = nn.ModuleList()
@@ -161,16 +154,12 @@ class ResBlock2(torch.nn.Module):
 
     def remove_weight_norm(self):
         for l in self.convs:
-            remove_weight_norm(l)
+            remove_parametrizations(l, "weight")
 
     def __prepare_scriptable__(self):
         for l in self.convs:
-            for hook in l._forward_pre_hooks.values():
-                if (
-                    hook.__module__ == "torch.nn.utils.weight_norm"
-                    and hook.__class__.__name__ == "WeightNorm"
-                ):
-                    torch.nn.utils.remove_weight_norm(l)
+            if is_parametrized(l, "weight"):
+                remove_parametrizations(l, "weight")
         return self
 
 
@@ -182,7 +171,7 @@ class ResidualCouplingLayer(nn.Module):
         kernel_size: int,
         dilation_rate: int,
         n_layers: int,
-        p_dropout: int = 0,
+        p_dropout: float = 0,
         gin_channels: int = 0,
         mean_only: bool = False,
     ):
@@ -249,12 +238,8 @@ class ResidualCouplingLayer(nn.Module):
         self.enc.remove_weight_norm()
 
     def __prepare_scriptable__(self):
-        for hook in self.enc._forward_pre_hooks.values():
-            if (
-                hook.__module__ == "torch.nn.utils.weight_norm"
-                and hook.__class__.__name__ == "WeightNorm"
-            ):
-                torch.nn.utils.remove_weight_norm(self.enc)
+        if is_parametrized(self.enc, "weight"):
+            remove_parametrizations(self.enc, "weight")
         return self
 
 
@@ -344,10 +329,6 @@ class ResidualCouplingBlock(nn.Module):
 
     def __prepare_scriptable__(self):
         for i in range(self.n_flows):
-            for hook in self.flows[i * 2]._forward_pre_hooks.values():
-                if (
-                    hook.__module__ == "torch.nn.utils.weight_norm"
-                    and hook.__class__.__name__ == "WeightNorm"
-                ):
-                    torch.nn.utils.remove_weight_norm(self.flows[i * 2])
+            if is_parametrized(self.flows[i * 2], "weight"):
+                remove_parametrizations(self.flows[i * 2], "weight")
         return self

rvc/layers/synthesizers.py

@@ -2,6 +2,7 @@ from typing import Optional, List, Union
 
 import torch
 from torch import nn
+from torch.nn.utils import parametrize
 
 
 from .encoders import TextEncoder, PosteriorEncoder
@@ -34,7 +35,7 @@ class SynthesizerTrnMsNSFsid(nn.Module):
         upsample_kernel_sizes: List[int],
         spk_embed_dim: int,
         gin_channels: int,
-        sr: Optional[Union[str, int]],
+        sr: Union[str, int],
         encoder_dim: int,
         use_f0: bool,
     ):
@@ -118,32 +119,16 @@ class SynthesizerTrnMsNSFsid(nn.Module):
             self.enc_q.remove_weight_norm()
 
     def __prepare_scriptable__(self):
-        for hook in self.dec._forward_pre_hooks.values():
-            # The hook we want to remove is an instance of WeightNorm class, so
-            # normally we would do `if isinstance(...)` but this class is not accessible
-            # because of shadowing, so we check the module name directly.
-            # https://github.com/pytorch/pytorch/blob/be0ca00c5ce260eb5bcec3237357f7a30cc08983/torch/nn/utils/__init__.py#L3
-            if (
-                hook.__module__ == "torch.nn.utils.weight_norm"
-                and hook.__class__.__name__ == "WeightNorm"
-            ):
-                torch.nn.utils.remove_weight_norm(self.dec)
-        for hook in self.flow._forward_pre_hooks.values():
-            if (
-                hook.__module__ == "torch.nn.utils.weight_norm"
-                and hook.__class__.__name__ == "WeightNorm"
-            ):
-                torch.nn.utils.remove_weight_norm(self.flow)
+        if parametrize.is_parametrized(self.dec, "weight"):
+            parametrize.remove_parametrizations(self.dec, "weight")
+        if parametrize.is_parametrized(self.flow, "weight"):
+            parametrize.remove_parametrizations(self.flow, "weight")
         if hasattr(self, "enc_q"):
-            for hook in self.enc_q._forward_pre_hooks.values():
-                if (
-                    hook.__module__ == "torch.nn.utils.weight_norm"
-                    and hook.__class__.__name__ == "WeightNorm"
-                ):
-                    torch.nn.utils.remove_weight_norm(self.enc_q)
+            if parametrize.is_parametrized(self.enc_q, "weight"):
+                parametrize.remove_parametrizations(self.enc_q, "weight")
         return self
 
-    @torch.jit.ignore
+    @torch.jit.ignore()
     def forward(
         self,
         phone: torch.Tensor,
@@ -155,18 +140,20 @@ class SynthesizerTrnMsNSFsid(nn.Module):
         pitchf: Optional[torch.Tensor] = None,
     ):  # 这里ds是id，[bs,1]
         # print(1,pitch.shape)#[bs,t]
-        g = self.emb_g(ds).unsqueeze(-1)  # [b, 256, 1]##1是t，广播的
+        embg = self.emb_g(ds).unsqueeze(-1)  # [b, 256, 1]##1是t，广播的
         m_p, logs_p, x_mask = self.enc_p(phone, pitch, phone_lengths)
-        z, m_q, logs_q, y_mask = self.enc_q(y, y_lengths, g=g)
-        z_p = self.flow(z, y_mask, g=g)
+        z, m_q, logs_q, y_mask = self.enc_q(y, y_lengths, g=embg)
+        z_p = self.flow(z, y_mask, g=embg)
         z_slice, ids_slice = rand_slice_segments_on_last_dim(
             z, y_lengths, self.segment_size
         )
-        if pitchf is not None:
+        if pitchf is not None and isinstance(self.dec, NSFGenerator):
             pitchf = slice_on_last_dim(pitchf, ids_slice, self.segment_size)
-            o = self.dec(z_slice, pitchf, g=g)
+            o = self.dec(z_slice, pitchf, g=embg)  # type: ignore
+        elif isinstance(self.dec, Generator):
+            o = self.dec(z_slice, g=embg)
         else:
-            o = self.dec(z_slice, g=g)
+            raise KeyError(f"unknown dec type: {type(self.dec).__name__}")
         return o, ids_slice, x_mask, y_mask, (z, z_p, m_p, logs_p, m_q, logs_q)
 
     @torch.jit.export
@@ -201,15 +188,17 @@ class SynthesizerTrnMsNSFsid(nn.Module):
             z_p = (m_p + torch.exp(logs_p) * torch.randn_like(m_p) * 0.66666) * x_mask
             z = self.flow(z_p, x_mask, g=g, reverse=True)
         del z_p, m_p, logs_p
-        if pitchf is not None:
+        if pitchf is not None and isinstance(self.dec, NSFGenerator):
             o = self.dec(
                 z * x_mask,
                 pitchf,
                 g=g,
                 n_res=return_length2,
             )
-        else:
+        elif isinstance(self.dec, Generator):
             o = self.dec(z * x_mask, g=g, n_res=return_length2)
+        else:
+            raise KeyError(f"unknown dec type: {type(self.dec).__name__}")
         del x_mask, z
         return o  # , x_mask, (z, z_p, m_p, logs_p)
 
@@ -326,7 +315,7 @@ class SynthesizerTrnMs256NSFsid_nono(SynthesizerTrnMsNSFsid):
         upsample_kernel_sizes: List[int],
         spk_embed_dim: int,
         gin_channels: int,
-        sr=None,
+        sr: Union[str, int],
     ):
         super().__init__(
             spec_channels,
@@ -346,6 +335,7 @@ class SynthesizerTrnMs256NSFsid_nono(SynthesizerTrnMsNSFsid):
             upsample_kernel_sizes,
             spk_embed_dim,
             gin_channels,
+            sr,
             256,
             False,
         )
@@ -371,7 +361,7 @@ class SynthesizerTrnMs768NSFsid_nono(SynthesizerTrnMsNSFsid):
         upsample_kernel_sizes: List[int],
         spk_embed_dim: int,
         gin_channels: int,
-        sr=None,
+        sr: Union[str, int],
     ):
         super().__init__(
             spec_channels,
@@ -391,6 +381,7 @@ class SynthesizerTrnMs768NSFsid_nono(SynthesizerTrnMsNSFsid):
             upsample_kernel_sizes,
             spk_embed_dim,
             gin_channels,
+            sr,
             768,
             False,
         )

rvc/layers/utils.py

@@ -1,4 +1,4 @@
-from typing import List, Optional, Tuple, Iterator
+from typing import List, Optional, Tuple, Iterator, Union
 
 import torch
 
@@ -17,7 +17,7 @@ def get_padding(kernel_size: int, dilation=1) -> int:
 
 def slice_on_last_dim(
     x: torch.Tensor,
-    start_indices: List[int],
+    start_indices: Union[List[int], torch.Tensor],
     segment_size=4,
 ) -> torch.Tensor:
     new_shape = [*x.shape]
@@ -32,9 +32,9 @@ def slice_on_last_dim(
 
 def rand_slice_segments_on_last_dim(
     x: torch.Tensor,
-    x_lengths: int = None,
+    x_lengths: Optional[Union[int, torch.Tensor]] = None,
     segment_size=4,
-) -> Tuple[torch.Tensor, List[int]]:
+) -> Tuple[torch.Tensor, Union[List[int], torch.Tensor]]:
     b, _, t = x.size()
     if x_lengths is None:
         x_lengths = t
@@ -58,7 +58,7 @@ def activate_add_tanh_sigmoid_multiply(
 def sequence_mask(
     length: torch.Tensor,
     max_length: Optional[int] = None,
-) -> torch.BoolTensor:
+):
     if max_length is None:
         max_length = int(length.max())
     x = torch.arange(max_length, dtype=length.dtype, device=length.device)

rvc/onnx/synthesizer.py

@@ -1,4 +1,4 @@
-from typing import List, Optional, Union
+from typing import List, Union
 
 import torch
 
@@ -25,7 +25,7 @@ class SynthesizerTrnMsNSFsid(SynthesizerBase):
         upsample_kernel_sizes: List[int],
         spk_embed_dim: int,
         gin_channels: int,
-        sr: Optional[Union[str, int]],
+        sr: Union[str, int],
         encoder_dim: int,
     ):
         super().__init__(

tools/web/infer-only.py

@@ -40,11 +40,9 @@ app = gr.Blocks()
 with app:
     with gr.Tabs():
         with gr.TabItem("在线demo"):
-            gr.Markdown(
-                value="""
+            gr.Markdown(value="""
                 RVC 在线demo
-                """
-            )
+                """)
             sid = gr.Dropdown(label=i18n("Inferencing voice"), choices=sorted(names))
             with gr.Column():
                 spk_item = gr.Slider(

web.py

@@ -36,7 +36,6 @@ import threading
 import shutil
 import logging
 
-
 logging.getLogger("numba").setLevel(logging.WARNING)
 logging.getLogger("httpx").setLevel(logging.WARNING)
 
@@ -89,23 +88,24 @@ index_paths = [""]
 
 
 def lookup_names(weight_root):
-    global names
+    names = []
     for name in os.listdir(weight_root):
         if name.endswith(".pth"):
             names.append(name)
+    return names
 
 
 def lookup_indices(index_root):
-    global index_paths
+    index_paths = []
     for root, _, files in os.walk(index_root, topdown=False):
         for name in files:
             if name.endswith(".index") and "trained" not in name:
                 index_paths.append(str(pathlib.Path(root, name)))
+    return index_paths
 
 
-lookup_names(weight_root)
-lookup_indices(index_root)
-lookup_indices(outside_index_root)
+names = [""] + lookup_names(weight_root)
+index_paths = [""] + lookup_indices(index_root) + lookup_indices(outside_index_root)
 uvr5_names = []
 for name in os.listdir(weight_uvr5_root):
     if name.endswith(".pth") or "onnx" in name:
@@ -113,12 +113,8 @@ for name in os.listdir(weight_uvr5_root):
 
 
 def change_choices():
-    global index_paths, names
-    names = [""]
-    lookup_names(weight_root)
-    index_paths = [""]
-    lookup_indices(index_root)
-    lookup_indices(outside_index_root)
+    names = [""] + lookup_names(weight_root)
+    index_paths = [""] + lookup_indices(index_root) + lookup_indices(outside_index_root)
     return {"choices": sorted(names), "__type__": "update"}, {
         "choices": sorted(index_paths),
         "__type__": "update",