Wave-U-Net

Cluster account setup

Follow step 1 listed here - https://isis.tu-berlin.de/pluginfile.php/2110087/mod_resource/content/16/python_cluster_tutorial.pdf

Environment setup (local and cluster)

We use miniconda as package manager. To install miniconda:

Download the installation script for Python 3.8 and your OS (download this for the cluster).
Run the installer: bash <script_name> and follow instructions on screen. When asked to run conda init, answer yes if you want conda to be initialized at login. That will add conda initialization in your .bashrc.
To ensure conda is initialized on ssh login to the cluster, you can call .bashrc from your .bash_profile like this:

if [ -f ~/.bashrc ]; then
  . ~/.bashrc
fi

The project environment is called pml and its dependencies are listed in environment.yml. Install all environment dependencies: conda env create -f environment.yml

We have to do this only once. Now we can always activate the conda environment with conda activate pml and deactivate it with conda deactivate.

Download the dataset using the bash script by running the following command. We only have to do this once. WARNING: THIS IS A 5GB DATASET.

cd data; source get_dataset.sh

#Usage