Can I track artifacts like confusion matrices or ROC curves?

Yes, the skill provides an artifact management API to save training plots, data versions, and environment files alongside your model metrics.

Does it work with my existing Weights & Biases setup?

Absolutely. The skill automatically detects W&B or MLflow in your project and configures them to log data in a way that remains tied to your development increments.

Can I use this without external tools like MLflow?

Yes, the skill includes a built-in tracking system that logs results locally to SpecWeave increment folders without any external dependencies.

How does it ensure experiment reproducibility?

It captures the git hash, data version, python environment, and exact hyperparameters for every run, ensuring anyone on the team can replicate the results.

How does this help with model documentation?

The skill integrates with SpecWeave's 'sync-docs' command to automatically update your internal documentation with experiment summaries and selection rationales.

ML Experiment Tracker

Name: ML Experiment Tracker
Author: anton-abyzov

byanton-abyzov

•

数据科学与机器学习

Manages machine learning experiment tracking and model comparison by automatically logging parameters, metrics, and artifacts to SpecWeave increments.

关于

The Experiment Tracker skill streamlines the machine learning development lifecycle by converting chaotic research into structured, reproducible experiments. It automatically detects and configures tracking backends like MLflow and Weights & Biases—or uses its own zero-config local storage—to ensure every model run is versioned alongside your code. By integrating directly with SpecWeave's living documentation, it transforms experimental data into persistent team knowledge, allowing developers to objectively compare models and document the rationale behind architectural decisions.

主要功能

Automatic logging of hyperparameters, metrics, and training artifacts
Comprehensive tracking of environment metadata, data versions, and git hashes
Built-in model comparison tool with automated markdown report generation
Synchronization of experiment results to project-wide living documentation
13 GitHub stars
Seamless integration with MLflow, Weights & Biases, and TensorBoard

使用场景

Tracking hyperparameter tuning results to identify optimal model configurations
Maintaining a reproducible history of experiments for team-wide knowledge sharing
Objective comparison of different ML architectures like XGBoost versus LightGBM

关于

主要功能

Automatic logging of hyperparameters, metrics, and training artifacts
Comprehensive tracking of environment metadata, data versions, and git hashes
Built-in model comparison tool with automated markdown report generation
Synchronization of experiment results to project-wide living documentation
13 GitHub stars
Seamless integration with MLflow, Weights & Biases, and TensorBoard

使用场景

Tracking hyperparameter tuning results to identify optimal model configurations
Maintaining a reproducible history of experiments for team-wide knowledge sharing
Objective comparison of different ML architectures like XGBoost versus LightGBM