Skip to content

Recipes#

Each recipe builds one real thing from start to finish: the choices, the steps in the console, with astra and with the API, what you should see, and how to check it. They need a workspace with a cluster and at least one machine (Add a machine), and the editor or admin role.

  • Develop in Cursor on your own GPU

    Your workstation's GPU as a development environment, opened in Cursor over Remote-SSH from your laptop, with no port opened.

    Uses: adding a machine, an environment pinned to it, astra ssh config, Open in Cursor.

  • From a notebook to an Astraeus training run

    Prototype in a notebook, write the script to the notebook's drive, and train it as a run on the same machine, reading the same files.

    Uses: a notebook, the notebooks drive, a run mounting it.

  • A team login node

    A shared, CPU-only, always-on environment where the team lands, keeps its scripts and submits runs to the GPUs.

    Uses: never idle (admins), sharing, a pool, astra astraeus run.

  • TensorBoard for a running training

    Train in an environment and watch the curves live in TensorBoard, at an address of its own.

    Uses: setup pip, an app on port 6006, astra env port.

  • Spark locally on a big machine

    Apache Spark in local mode on all the cores you ask for, its scratch space on the drive, and its UI in your browser.

    Uses: the spark preset, spark-submit, pyspark, an app on 4040.