maxschnupp/BADDSP

Code related to my "Bachelor Arbeit" on DDSP

BUILDING:

sudo docker build -t baddsp ./

RUNNING/OPENNING:

sudo docker run -it baddsp

TRAINING:

• When restarting an instance you will have to reinstall docker before moving on to the symlink step

• Start a GPU instance on GCS and attach a disk

• Connect by SSH and mount https://cloud.google.com/compute/docs/disks/add-persistent-disk

• Change docker image installation registry to disk with a symlink https://forums.docker.com/t/how-do-i-change-the-docker-image-installation-directory/1169

• upload serviceKey.json to /app

• build: sudo docker build -t baddsp ./

• run: sudo docker run -it baddsp

• train: python3 main.py --train instrument-name

TRANSFER:

• input files should be put in the audioIn folder after transfer they will be available in audioOut and also in the bucket

• single file: python3 main.py --transfer source-file-name.wav output-file-name.wav label the label will be prepended to the file name in the bucket

• all files python3 main.py --transferAll label will iterate through the entire directory and transfer timbre for all files