Installation

pip install nb2slurm

Python 3.9 or newer. The dependencies (papermill, filelock, jinja2, paramiko) are installed with it.

What you need besides the package

On your own machine (laptop, or a Jupyter server):

  • nb2slurm and Jupyter, to run your workflow for one subject and drive the cluster.

  • rsync, used by push() and pull(). On Windows use WSL, Git Bash or conda’s rsync — the WSL notebook installs one for you.

  • SSH access to the cluster — a key is easiest:

    import nb2slurm
nb2slurm.generate_key(key_type="ed25519")          # only once, if you have no key yet
print(nb2slurm.public_key("~/.ssh/id_ed25519"))    # paste this into your HPC account

    See Connecting to the cluster for registering the key, passphrases and test_connection().

On the cluster:

  • nb2slurm installed in the environment the jobs run in, and a registered Jupyter kernel for papermill to execute the notebooks with. nb2slurm can create both for you — see Environments and kernels.

  • rclone, only if you use mounts=[...].

Workflow.check() verifies the cluster side for you before the first submit.

Development install

Working on nb2slurm itself rather than using it? The clone, test, lint and docs-build steps live in the README.