https://github.com/mddct/transformer-vocos
https://github.com/mddct/transformer-vocos
Last synced: 5 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/mddct/transformer-vocos
- Owner: Mddct
- Created: 2025-03-13T09:13:43.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2025-03-13T15:37:02.000Z (over 1 year ago)
- Last Synced: 2025-03-13T16:35:16.379Z (over 1 year ago)
- Size: 13.7 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
## ongoing
### Why vocos with transformer or conformer ?
Easy to scale and Good control over latency and caching
### Why sequence mask in gan?
No data length limit is required, such as 1s
### Why in wenet ？
cache and multiple speech models are available out of the box
### Data Prepare
```bash
{"wav": "/data/BAC009S0764W0121.wav"}
{"wav": "/data/BAC009S0764W0122.wav"}
```
### train
```bash
train_data = 'train.jsonl'
model_dir = 'vocos/exp/2025/0.1/transformer/'
tensorboard_dir = ${model_dir}/runs/
mkdir -p $model_dir $tensorboard_dir
torchrun --standalone --nnodes=1 --nproc_per_node=8 vocos/main.py -- \
--config vocos/configs/default.py \
--config.train_data=${train_data} \
--config.model_dir=${model_dir} \
--config.tensorboard_dir=${tensorboard_dir} \
--config.max_train_steps 1000000
```
TODO:
- [ ] training
- [x] training works
- [x] check training process
- [x] generator
- [x] disc
- [ ] distill
- [x] resume
- [ ] stereo for music
- [x] cqt loss
- [ ] dev benchmark etc
- [ ] infer
- [x] offline
- [ ] chunk by chunk or frame by frame
- [ ] onnx
- [ ] exmple for: cosyvoice2 and transformer-vocos