HPC tutorials#

Estimated reading time:2 minutes

Reference setup for running EEGDash on shared HPC clusters: SLURM submission scripts (CPU and GPU), a Dockerfile, and a tutorial showing how to combine the on-disk cache with batch scheduling for an eyes-open / eyes-closed run. Difficulty 2; assumes the core workflow track and that you have admin access on a SLURM cluster.

Single-machine examples cover the API; the HPC track covers the operational layer that lets you run those examples at the scale a real study requires. The cache layout, environment setup, and SLURM array patterns here are the same ones used in production for the EEGDash benchmark runs. Sourced from docs/tutorial_restructure_plan.md Category J (HPC operations). Cross-link with the How-To guide for parallel feature extraction.

What you will learn:

How to package an EEGDash environment in a Docker image so cluster jobs see the same dependencies you tested locally.
How to write a SLURM submission script that pins one EEGDash job per GPU (run_eoec_gpu.slurm) or per CPU node (run_eoec_cpu.slurm).
How to lay out the on-disk EEGDash cache so concurrent jobs share preprocessed windows without race conditions.
How to run an eyes-open / eyes-closed end-to-end study from a single sbatch invocation, with logs and outputs landing in the right cache location for downstream analysis.

Each script in this directory is a runnable example; instructions.md walks through the cluster-side setup once before you launch.

Eyes Open vs. Closed Classification