mirror of
https://github.com/w-okada/voice-changer.git
synced 2025-02-02 16:23:58 +03:00
56 lines
1.5 KiB
Markdown
56 lines
1.5 KiB
Markdown
|
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
|
||
|
|
```
|