See Trackio: Free, native, open source Python library for experimental trackers to simplify and enhance machine learning workflows

Experimental tracking is an important part of modern machine learning workflows. Whether you are adjusting hyperparameters, monitoring training metrics, or working with colleagues, you must have robust, flexible tools to make tracking experiments direct and insightful. However, many existing experimental tracking solutions require complex setups, with licensing fees or locking user data into proprietary formats, making them less accessible to individual researchers and smaller teams.

Meet Trackio – A new open source experimental tracking library developed by Hug Face and Gradio. Trackio is a local, lightweight and completely free tracker designed for today’s fast-paced research environment and open collaboration.

What is Trackio?

Trackio is a python package designed as Replacement replacement For widely used libraries like WandB, with compatibility with underlying API calls (wandb.init,,,,, wandb.log,,,,, wandb.finish). This puts Trackio into a coalition where old scripts toggle or run require little change of code, which in short is like importing Trackio as Wandb and continuing to work as before.

Key Features

• The first local design: By default, experiments run locally and persist, providing privacy and quick access. Sharing is optional, not the default value.
• Free and open source: There is no paywall, no feature limitations, and anything including collaboration and online dashboards are available for free.
• Lightweight and scalable: Python lines for the entire code base are below 1,000 lines, ensuring ease of auditing, scaling, or adapting.
• Integrate with the embracing facial ecosystem: Out-of-the-box support Transformers,,,,, Sentence Transformersand Acceleratelet users start using the least settings to track metrics.
• Data portability: Unlike some established tracking tools, Trackio makes all experimental data easy to export and accessible, giving custom analysis and seamless integration into the research pipeline.

Seamless experiment tracking: local or shared

One of the great features of Trackio is Shareability. Researchers can monitor metrics on local-level driven dashboards, or by simply synchronizing with the embracing facial space, they can migrate the dashboards online to share with colleagues (or the public if needed). Spaces can be set to private or public places – no complicated authentication or the boarding job required by the audience.

For example, view your experiment dashboard locally:

Or, from Python:

import trackio
trackio.show()

Launch the dashboard on space:

• Synchronize logs to the embracing facial space And share or embed the experiment dashboard immediately with simple URLs.

Importantly, when running on space, Trackio automatically backs up every 5-minute metrics from a brief SQLite DB to a hugged face dataset (such as a parquet file), ensuring that your experimental data is never lost – even if the public space restarts.

Plugin integration with your ML workflow

Integration with the embrace face ecosystem is crucial:

• and transformers.Trainer or accelerateyou can record and visualize metrics by specifying Trackio as a logger.

For example, use acceleration:

from accelerate import Accelerator
accelerator = Accelerator(log_with="trackio")
accelerator.init_trackers("my-experiment")
...
accelerator.log({"training_loss": loss}, step=step)

This low friction method refers to experiments where anyone using a transformer, sentence transformer or acceleration can immediately start tracking and sharing zero additional settings.

Transparency, sustainability and data freedom

Trackio goes a step further than standard metrics, encouraging transparency in the use of computing resources. It supports tracking metrics GPU energy use (by reading nvidia-smi), this feature is in line with the emphasis on environmental responsibility and repeatability in model card documents.

Unlike closed platforms, Your data is always accessibleTrackio’s logs are stored in a standard format, and the dashboard is built using open tools such as Gradio and hug the face datasets, making everything easy to mix, analyze or share.

Start quickly

start:

pip install trackio
# or
uv pip install trackio

Or, swap imports in your codebase:

in conclusion

Trackio Positioning to authorize Individual researchers and public collaborations Provide transparent and completely free experiment tracker in ML. By default, local priority, easy to share, and tightly integrated with embracing facial tools, it brings a promise of strong tracking without the friction or cost of traditional solutions.

Check Technical details and Github page. Check out ours anytime Tutorials, codes and notebooks for github pages. Also, please stay tuned for us twitter And don’t forget to join us 100K+ ml reddit And subscribe Our newsletter.