minrecord

Overview

minrecord is a minimalist Python library to record values in a ML workflow. In particular, it provides functionalities to track the best value, or the most recent values by storing a limiting number of values. It is possible to customize the library e.g. it is possible to define a new logic to track the best value.

Key Features

• 📊 Track Best Values: Automatically track the best value seen during training
• 🔄 Recent History: Store only recent values to minimize memory usage
• ✅ Improvement Detection: Easily check if your model is still improving
• 🎯 Flexible Comparators: Define custom logic to determine what "better" means
• 🗂️ Record Manager: Organize and manage multiple metrics efficiently
• 💾 Serialization: Save and load record states for checkpointing
• 🚀 Minimal Dependencies: Lightweight with only essential dependencies

Below is an example to show how to track the best scalar value when the best value is the maximum value and when the best value is the minimum value.

>>> from minrecord import MaxScalarRecord, MinScalarRecord
>>> record_max = MaxScalarRecord("accuracy")
>>> record_max.update([(0, 42), (None, 45), (2, 46)])
>>> record_max.add_value(40)
>>> record_max.get_best_value()
46
>>> record_min = MinScalarRecord("loss")
>>> record_min.update([(0, 42), (None, 45), (2, 46)])
>>> record_min.add_value(50)
>>> record_min.get_best_value()
42

Documentation

• latest (stable): documentation from the latest stable release.
• main (unstable): documentation associated to the main branch of the repo. This documentation may contain a lot of work-in-progress/outdated/missing parts.

Installation

We highly recommend installing a virtual environment. minrecord can be installed from pip using the following command:

pip install minrecord

To make the package as slim as possible, only the minimal packages required to use minrecord are installed. To include all the dependencies, you can use the following command:

pip install minrecord[all]

Please check the get started page to see how to install only some specific dependencies or other alternatives to install the library. The following is the corresponding minrecord versions and tested dependencies.

^* indicates an optional dependency

Contributing

Please check the instructions in CONTRIBUTING.md.

Suggestions and Communication

Everyone is welcome to contribute to the community. If you have any questions or suggestions, you can submit Github Issues. We will reply to you as soon as possible. Thank you very much.

API stability

:warning: While minrecord is in development stage, no API is guaranteed to be stable from one release to the next. In fact, it is very likely that the API will change multiple times before a stable 1.0.0 release. In practice, this means that upgrading minrecord to a new version will possibly break any code that was using the old version of minrecord.

License

minrecord is licensed under BSD 3-Clause "New" or "Revised" license available in LICENSE file.