WIP: v1.5support 1

docker_trainer/Dockerfile

@@ -0,0 +1,39 @@
+FROM nvidia/cuda:11.8.0-cudnn8-runtime-ubuntu22.04 as base
+
+ARG DEBIAN_FRONTEND=noninteractive
+
+RUN apt-get -y update && apt-get install -y emacs git curl python3-pip libsndfile1-dev ffmpeg
+
+RUN git clone --depth 1 -b v1.5.0.0_SiFiGAN https://github.com/isletennos/MMVC_Trainer.git
+
+ADD /model/D_v15_best.pth /MMVC_Trainer/fine_model/
+ADD /model/G_v15_best.pth /MMVC_Trainer/fine_model/
+
+RUN pip install pydub==0.25.1
+RUN pip install tqdm==4.64.1
+RUN pip install librosa==0.9.2
+RUN pip install torch==1.13.1
+RUN pip install torchaudio==0.13.1
+RUN pip install psutil==5.9.4
+RUN pip install tensorboard==2.11.2
+RUN pip install pytz==2022.7.1
+RUN pip install pyworld==0.3.2
+RUN pip install retry==0.9.2
+RUN pip install h5py==3.8.0
+RUN pip install matplotlib==3.6.3
+
+ADD /warmup.py /
+RUN python3 warmup.py
+WORKDIR /MMVC_Trainer/
+
+# python3 normalize.py True
+# python3 create_dataset.py -f train_config -s 24000 -m dataset/multi_speaker_correspondence.txt
+# tensorboard --logdir logs --port 5000
+# python3 train_ms.py -c configs/train_config.json -m 20220306_24000 -fg fine_model/G_v15_best.pth -fd fine_model/D_v15_best.pth
+
+#$ docker run --gpus all --rm -ti -v `pwd`/trainer/dataset:/MMVC_Trainer/dataset -v `pwd`/trainer/configs:/MMVC_Trainer/configs -v `pwd`/trainer/F0:/MMVC_Trainer/F0 -v `pwd`/trainer/cF0:/MMVC_Trainer/cF0 -v `pwd`/trainer/units:/MMVC_Trainer/units -v `pwd`/trainer/logs:/MMVC_Trainer/logs -v `pwd`/trainer/filelists:/MMVC_Trainer/filelists -p 5000:5000 trainer /bin/bash
+
+
+
+
+

docker_trainer/README.md

@@ -0,0 +1,55 @@
+MMVC Server
+----
+# 起動方法
+
+(1) Datasetを`trainer/dataset`におく
+```sh
+trainer/dataset/
+├── 00_myvoice
+│   ├── text
+│   │   ├── emotion001.txt
+│   │   ├── emotion002.txt
+...
+│   │   └── emotion100.txt
+│   └── wav
+│       ├── emotion001.wav
+│       ├── emotion002.wav
+...
+│       └── emotion100.wav
+├── 1205_zundamon
+│   ├── text
+│   │   ├── emoNormal_001.txt
+│   │   ├── emoNormal_002.txt
+...
+│   │   └── emoNormal_100.txt
+│   └── wav
+│       ├── emoNormal_001.wav
+│       ├── emoNormal_002.wav
+...
+│       └── emoNormal_100.wav
+├── 344_tsumugi
+│   ├── text
+│   │   ├── VOICEACTRESS100_001.txt
+│   │   ├── VOICEACTRESS100_002.txt
+...
+│   │   └── emoNormal_100.txt
+│   └── wav
+│       ├── VOICEACTRESS100_001.wav
+│       ├── VOICEACTRESS100_002.wav
+...
+│       └── emoNormal_100.wav
+└── multi_speaker_correspondence.txt
+```
+
+(2) start.shをrootにコピー
+
+(3) `bash start.sh`を実行
+
+(4) Docker内で次のコマンドを実行
+batch sizeは適宜調整
+```sh
+$ python3 normalize.py True
+$ python3 create_dataset.py -f train_config -s 24000 -m dataset/multi_speaker_correspondence.txt
+$ tensorboard --logdir logs --port 5000
+$ python3 train_ms.py -c configs/train_config.json -m 20220306_24000 -fg fine_model/G_v15_best.pth -fd fine_model/D_v15_best.pth
+```

docker_trainer/model/.gitignore

@@ -0,0 +1,2 @@
+*
+!.gitignore

docker_trainer/start.sh

@@ -0,0 +1,14 @@
+docker run --gpus all --rm -ti \
+    -v `pwd`/trainer/dataset:/MMVC_Trainer/dataset \
+    -v `pwd`/trainer/configs:/MMVC_Trainer/configs \
+    -v `pwd`/trainer/F0:/MMVC_Trainer/F0 \
+    -v `pwd`/trainer/cF0:/MMVC_Trainer/cF0 \
+    -v `pwd`/trainer/units:/MMVC_Trainer/units \
+    -v `pwd`/trainer/logs:/MMVC_Trainer/logs \
+    -v `pwd`/trainer/filelists:/MMVC_Trainer/filelists \
+    -p 5000:5000 \
+    trainer /bin/bash
+
+
+
+

docker_trainer/warmup.py

@@ -0,0 +1,3 @@
+import torch
+
+hubert = torch.hub.load("bshall/hubert:main", "hubert_soft")

package.json

@@ -6,6 +6,7 @@
   "scripts": {
     "build:docker": "date +%Y%m%d%H%M%S > docker/dummy && DOCKER_BUILDKIT=1 docker build -f docker/Dockerfile docker/ -t voice-changer",
     "build:docker:onnx": "DOCKER_BUILDKIT=1 docker build -f docker_onnx/Dockerfile docker/ -t onnx-converter",
+    "build:docker:trainer": "DOCKER_BUILDKIT=1 docker build -f docker_trainer/Dockerfile docker_trainer/ -t trainer",
     "copy:frontend": "docker run -v `pwd`/frontend/dist:/frontend/dist --entrypoint /bin/bash -ti voice-changer -c \"cp -r /voice-changer-internal/frontend/dist/* /frontend/dist\"",
     "copy:backend": "docker run -v `pwd`/demo:/demo --entrypoint /bin/bash -ti voice-changer -c \"cp -r /voice-changer-internal/voice-change-service/* /demo/ && rm -rf /demo/MMVC_Trainer/.git && rm -rf /demo/MMVC_Trainer/.gitignore \"",
     "create:demo": "run-p copy:frontend copy:backend",

trainer/F0/.gitignore

@@ -0,0 +1,2 @@
+*
+!.gitignore

trainer/cF0/.gitignore

@@ -0,0 +1,2 @@
+*
+!.gitignore

trainer/configs/.gitignore

@@ -0,0 +1,2 @@
+*
+!.gitignore

trainer/dataset/.gitignore

@@ -0,0 +1,2 @@
+*
+!.gitignore

trainer/logs/.gitignore

@@ -0,0 +1,2 @@
+*
+!.gitignore

trainer/units/.gitignore

@@ -0,0 +1,2 @@
+*
+!.gitignore