█

AI/TLDR

Buy me a coffee
Buy me a coffee

Aim

Self-hosted, open-source experiment tracker built for 10,000s of training runs

Experiment TrackingOpen source
Language
Python
License
Apache-2.0
$pip3 install aim
IN SHORT

What is Aim?

Aim is an open-source, self-hosted tool for tracking machine learning experiments. It sits in the Experiment Tracking category of the Training & Fine-Tuning tools and is open-source, written in Python and licensed Apache-2.0. To get started: install Aim.

Overview

Aim is an open-source, self-hosted tool for tracking machine learning experiments. You log training runs and any related metadata from your code, then explore and compare them in a web UI. It is designed to handle tens of thousands of runs without slowing down.

It fits the experiment-tracking corner of MLOps: the place where you keep a record of every training run, its hyperparameters, and its metrics so you can see what worked. Because Aim runs on your own machine or server, you keep full control of your data instead of sending it to a hosted service.

It is aimed at ML engineers and researchers who run many experiments and want one dashboard to group, filter, and compare them. Alongside the UI, a Python SDK lets you query tracked metadata programmatically, which is handy for automation and analysis inside Jupyter notebooks.

What it does

  • Logs training runs and arbitrary AI metadata, with integrations for popular ML frameworks
  • Web UI to visualize, group, aggregate, and compare runs across experiments
  • Query runs using Python expressions, both in the UI and through the SDK
  • Tracks system info and resource usage during training
  • Organize runs with tags and experiments plus a centralized dashboard
  • Self-hosted, so your experiment data stays on your own infrastructure

Getting started

Install Aim with pip, add a few lines to your training script to track metrics, then launch the UI to compare runs.

Install Aim

Install the package from PyPI.

bashbash
pip3 install aim

Track a run in your code

Create a Run, log your hyperparameters, and track metrics in your training loop.

pythonpython
from aim import Run

# Initialize a new run
run = Run()

# Log run parameters
run["hparams"] = {
    "learning_rate": 0.001,
    "batch_size": 32,
}

# Log metrics
for i in range(10):
    run.track(i, name='loss', step=i, context={ "subset":"train" })
    run.track(i, name='acc', step=i, context={ "subset":"train" })

Open the UI

Start the Aim web interface to explore and compare your tracked runs.

bashbash
aim up

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

When to use it

  • Compare hyperparameter sweeps across many runs to find the configuration that trains best
  • Keep a self-hosted record of all experiments on your own server instead of a third-party service
  • Query tracked metrics from a Jupyter notebook for custom analysis or reporting
  • Monitor metrics and system resource usage while a training job is running

How Aim compares

Aim alongside other open-source experiment tracking tools AI/TLDR tracks, ranked by GitHub stars.

ToolStarsWhat it does
MLflow★ 28.9kAn open-source ML and GenAI platform that, alongside experiment tracking, provides evaluation and monitoring tools for scoring and comparing LLM and agent outputs.
DVC★ 16kA Git-like version control tool for datasets, models, and experiments that makes ML projects reproducible.
Weights & Biases (wandb)★ 11.3kAn open-source client library for logging experiments, running hyperparameter sweeps, and versioning models and datasets.
ClearML★ 6.9kAn open-source MLOps suite whose experiment manager auto-logs runs, metrics, and artifacts and ties them into data management and orchestration.
Aim★ 6.3kSelf-hosted, open-source experiment tracker built for 10,000s of training runs
Sacred★ 4.4kA Python tool for configuring, organizing, logging, and reproducing computational research experiments.
Polyaxon★ 3.7kAn open-source platform for tracking, scheduling, and managing large-scale deep learning experiments on Kubernetes.
Trackio★ 1.8kA lightweight open-source experiment tracker from Hugging Face that stores runs locally in SQLite as a drop-in alternative to wandb.