Manage machine learning experiment lifecycles from planning to reporting.
npx skills add https://github.com/ericmjl/skills --skill ml-experimentation
Please help me install this Agent Skill. Skill: ml-experimentation Source: https://github.com/ericmjl/skills/tree/main/skills/ml-experimentation Command: npx skills add https://github.com/ericmjl/skills --skill ml-experimentation
Planning, executing, and reporting machine learning experiments can be complex and error-prone; this skill provides a repeatable lifecycle with journaling, diagnostics, and a canonical artifact structure to improve reproducibility and communication.
Plan a new hypothesis, initialize an experiment, and run an initial de-risked iteration using the provided scripts to generate a JOURNAL.md and baseline logs.
High-intent search queries and answers about installing and using this skill.
To manage the end-to-end machine learning experiment lifecycle, you can use a structured workflow that supports hypothesis-driven iteration, journaling, plotting, and formal report writing. This ensures reproducibility through a canonical runs structure including JOURNAL.md, logs, and plots.
To structure ML experiment runs for reproducibility and auditability, maintain a strict canonical tree with JOURNAL.md, runs/, logs/, and plots/. This deterministic artifact structure enables fast iteration, scientific reporting, and clear tracking of log-based metrics.
Yes, you need a deterministic workflow with uv or pixi for script execution when running hypothesis-driven ML experiments. This requirement ensures that your de-risked runs and full validation iterations execute consistently and reliably across different environments.
The best way to run rapid de-risked iterations for machine learning experiments is to execute scripts under 60 seconds that generate baseline logs and a JOURNAL.md. This approach allows quick hypothesis testing before committing to full validation runs.
Journaling during fast ML experiment iterations works by recording observations directly in a JOURNAL.md file within a canonical runs structure. This practice captures hypothesis changes, logs metrics, and supports plotting and final scientific report generation.
Yes, you can use this approach for full hypothesis validation runs. The workflow supports both rapid de-risked runs under 60 seconds and full runs for comprehensive hypothesis validation, utilizing structured data, logs, and plots for final reporting.
KameniAlexNeaManage Python virtual environments and dependencies with uv.
ovachieverGenerate publication-ready scientific figures with matplotlib, seaborn, and plotly.
zakura-coreGenerate SVG plots and diagnostic summaries from Zebra Zakura performance trace directories.
doanchienthangdevAutomate machine learning experiment design, baseline establishment, and tracking with MLflow.
