autogluon/autogluon

★ 10,326PythonAudience · dataComplexity · 3/5Setup · moderate

Mindmap

mindmap
  root((AutoGluon))
    What it does
      Auto model training
      Ensemble stacking
      Multi-data-type support
    Use Cases
      Tabular prediction
      Time series forecast
      Multimodal modeling
    Tech Stack
      Python
    Audience
      Data practitioners
      ML beginners
      Business analysts

mindmap root((AutoGluon)) What it does Auto model training Ensemble stacking Multi-data-type support Use Cases Tabular prediction Time series forecast Multimodal modeling Tech Stack Python Audience Data practitioners ML beginners Business analysts

Click or tap to explore — scroll the page freely

Things people build with this

USE CASE 1

Train a predictive model on a CSV of business data without picking an algorithm or adjusting any settings.

USE CASE 2

Build a time series forecasting model for daily sales or sensor readings with minimal code.

USE CASE 3

Create a model that accepts mixed text, images, and numbers to predict outcomes in a multimodal dataset.

Tech stack

Python

Getting it running

Difficulty · moderate Time to first run · 30min

Requires Python 3.10 or newer, GPU support is optional and not needed for basic tabular tasks.

In plain English

AutoGluon is a Python library, built by the AWS AI team, that lets you train machine learning models without needing to understand the details of how those models work internally. Machine learning normally requires a lot of manual tuning: picking which algorithm to use, adjusting dozens of settings, and combining multiple models together. AutoGluon handles all of that automatically, so you can go from a data file to a working predictive model in three lines of code. The library covers four main types of data. Tabular data means spreadsheet-style rows and columns, which is the most common format for business data. Time series data means measurements recorded over time, like daily sales figures or hourly sensor readings. Multimodal data means inputs that mix text, images, and numbers together in a single prediction task. AutoGluon figures out how to process each type and picks an approach that tends to perform well without you needing to specify one. Under the hood, AutoGluon trains many different models on your data and then combines their predictions using a stacking technique, where one model learns to correct the mistakes of others. This ensemble approach is why it tends to score well in data science competitions and benchmarks. The README links to several academic papers published at venues like NeurIPS and ICML that document this approach in detail. Installing it requires Python 3.10 or newer and works on Linux, macOS, and Windows. You install it with a single pip command. The documentation site includes quickstart tutorials, a full API reference, and guides for each data type. There is also a Discord community and the project posts release notes as it adds new features. The library is aimed at developers and data practitioners who want accurate predictions without investing significant time in model selection or tuning. It does not replace understanding what your data means or whether a model's output makes sense for your use case, but it removes most of the technical trial and error that normally comes before getting a working result.

Copy-paste prompts

Prompt 1

Using AutoGluon, write the Python code to train a tabular predictor on a CSV file and print its accuracy on a test set.

Prompt 2

How do I use AutoGluon's TimeSeriesPredictor to forecast the next 7 days of sales from a daily sales CSV?

Prompt 3

Show me how to train an AutoGluon multimodal model that takes both a product image and a text description to predict a category.

Prompt 4

How does AutoGluon's stacking ensemble work, and how do I inspect which base models it selected?

Prompt 5

What pip command installs AutoGluon on macOS with Python 3.10, and how long does a basic tabular fit typically take?

Open on GitHub → Explain another repo

← autogluon on gitmyhub — every repo by this author, as a profile.

Verify against the repo before relying on details.