h1de0us/source-separation

Installation guide

To install all dependencies, run:

pip install -r ./requirements.txt

To create a mix, run ./create_mixture.sh to download the dataset automatically and create mixes with the default parameters, or download dataset manually and run the following command:

python3 mix-creation.py \
    --speakers_files_train <your-path-to-train-data> \
    --speakers_files_val <your-path-to-val-data> \
    --path_mixtures_train <output-folder-for-training-mixes> \
    --path_mixtures_val <output-folder-for-validation-mixes> \
    --nfiles-train <number-of-files-to-use-for-training-mixes> \
    --nfiles-val <number-of-files-to-use-for-validation-mixes>

To (try to) train a model, run

python3 train.py -c hw_ss/config.json

Unfortunately, there is no checkpoint, as all my attempts to train the model were unsuccessful. However, I was very close :(

Credits

This repository is based on a heavily modified fork of pytorch-template repository.

Docker

You can use this project with docker. Quick start:

docker build -t my_hw_ss_image . 
docker run \
   --gpus '"device=0"' \
   -it --rm \
   -v /path/to/local/storage/dir:/repos/asr_project_template/data/datasets \
   -e WANDB_API_KEY=<your_wandb_api_key> \
	my_hw_asr_image python -m unittest 

Notes:

• -v /out/of/container/path:/inside/container/path -- bind mount a path, so you wouldn't have to download datasets at the start of every docker run.
• -e WANDB_API_KEY=<your_wandb_api_key> -- set envvar for wandb (if you want to use it). You can find your API key here: https://wandb.ai/authorize