marimo-batch

The marimo-batch skill is designed to streamline the preparation of marimo notebooks for scheduled batch runs, primarily targeting machine learning workflows. It provides a structured and opinionated framework for defining parameters, environment configurations, and execution patterns, ensuring consistency and reproducibility in batch jobs. By integrating with both a graphical user interface and the command-line interface, marimo-batch allows users to efficiently configure, test, and execute notebooks with minimal setup, addressing the common challenge of managing complex notebook runs and parameter configurations for ML experiments.

Key features include the ability to define model parameters using Pydantic, providing a single source of truth for hyperparameters such as sample size and learning rate. Users can configure these parameters via an interactive UI form or through the CLI, offering flexibility between rapid experimentation and automated batch execution. The skill also supports environment variable management using python-dotenv and EnvConfig, allowing users to securely handle API keys and other configuration values. Additional features include support for column-based notebook organization, logging best practices for monitoring batch jobs, and guidance for integrating tools like Weights and Biases for tracking experiment metrics. Users are also guided on maintaining notebook integrity and performing grid search or hyperparameter sweeps.

The marimo-batch skill is ideal for data scientists, ML engineers, and researchers who need a robust way to prepare and execute notebooks in a reproducible and configurable manner. It is particularly useful for those who want to iterate quickly using a UI for experimentation while also leveraging the CLI for scheduled or large-scale batch jobs. Target users benefit from the skill's structured approach to parameter management, environment setup, and logging, enabling faster iteration, easier debugging, and consistent execution of machine learning workflows.