Auto-PyTorch
automl.github.io
- Free plan
- Linux
- Self-hosted
What Auto-PyTorch is
Auto-PyTorch is a free, self-hosted machine-learning toolkit based on PyTorch that automates model selection and hyperparameter tuning. It searches for model configurations within a user-set budget, using Bayesian optimization, meta-learning, SMAC and Hyperband; neural-network architecture and training hyperparameters can be optimized together. Supported tasks include tabular classification, tabular regression and time-series forecasting. For tabular work, preprocessing can cover imputation, categorical encoding, scaling and feature processing, with related hyperparameters tuned during search. The toolkit can assemble ensembles using models' predictions on a validation set, and users can configure ensemble size and candidate limits. Memory limits for estimators and an overall wall-time limit are configurable. Parallel Bayesian optimization uses Dask.distributed, but workers need shared filesystem access to training data and models. The project provides a Docker image and documents PyPI or manual installation. Requirements include Linux, Python 3.7 or later, a C++11-capable compiler and SWIG 3.0; SWIG 4.0 or later is not supported. Forecasting needs extra dependencies.
Who it suits
It suits developers and machine-learning practitioners who can work in a self-hosted Python environment and want automated model search for supported tabular or forecasting tasks.
Good to know, and what to watch
- Automates model selection and hyperparameter tuning.
- Supports classification, regression and time-series forecasting.
- Users can set estimator memory and search time limits.
- Available through PyPI, manual installation and Docker.
- Installation requirements include Linux and SWIG 3.0.
- SWIG 4.0 or later is not supported.
- Forecasting requires additional dependencies.
- Parallel workers require access to a shared file system.
Plans
Every plan and priceAuto-PyTorch
Free plan
Getting started
- Open the Auto-PyTorch documentation website
- Use Linux with Python 3.7 or later
- Install a C++11-capable compiler and SWIG 3.0
- Install from PyPI or manually in a Python environment
- Alternatively use the documented Docker image
- Add the extra dependencies required for forecasting when needed
Limits to check
The software is free, but installation has Linux, Python, compiler and SWIG requirements. Parallel workers need shared filesystem access, and time-series forecasting needs dependencies beyond the base installation.
The verdict
Auto-PyTorch offers automated search and ensembling for several task types, with resource controls and parallel processing options. Check its Linux and SWIG requirements, and plan for extra forecasting dependencies if needed.
Questions people ask
Is Auto-PyTorch free?
Yes. The Auto-PyTorch plan is free, and the project uses the 3-clause BSD license.
Which platforms does it support?
The listed platforms are Linux and self-hosted environments.
What machine-learning tasks are supported?
Documentation covers tabular classification, tabular regression and time-series forecasting.
How can it be installed?
The project documents PyPI installation, manual installation in a Python environment and a Docker image.
Can searches run in parallel?
Yes, parallel Bayesian optimization is available through Dask.distributed. Workers need shared filesystem access to training data and models.
Who develops Auto-PyTorch?
The GitHub project identifies the AutoML Groups of the University of Freiburg and Hannover as its developers.
What the maker publishes
- Compatibility limit
- The installation documentation says SWIG 4.0 or later is not supported. source
- Data preparation
- For tabular tasks, its preprocessing includes imputation, categorical encoding, scaling, and feature preprocessing, with corresponding hyperparameters tuned during search. source
- Deployment
- The project provides a Docker image and can be installed from PyPI or manually in a Python environment. source
- Ensembling
- It builds ensembles by selecting among models based on their predictions for a validation set, and users can configure ensemble size and candidate limits. source
- Forecasting dependencies
- Time-series forecasting requires additional dependencies beyond the base installation. source
- Installation
- The installation documentation specifies Linux, Python 3.7 or later, a C++11-capable compiler, and SWIG 3.0.*, and also documents a Docker image. source
- Integration
- The documented ecosystem includes PyTorch, scikit-learn transformers, Dask.distributed, and threadpoolctl. source
- Integrations
- The documentation describes using Dask.distributed for parallel Bayesian optimization and sklearn column transformers for data preprocessing. source
- License
- The documentation states that Auto-PyTorch is licensed under the 3-clause BSD license. source
- Maker
- The GitHub project says Auto-PyTorch is developed by the AutoML Groups of the University of Freiburg and Hannover. source
- Optimization
- It jointly optimizes neural network architecture and training hyperparameters for automated deep learning. source
- Optimization methods
- It uses Bayesian optimization, meta-learning, and ensemble construction to search for models. source
- Parallel computing requirement
- When using multiple workers, the documentation says they must have access to a shared file system for training data and models. source
- Parallel processing
- It supports parallel Bayesian optimization using Dask.distributed, and parallel workers need access to a shared file system for training data and models. source
- Platform requirement
- The installation documentation lists Linux, Python 3.7 or later, a C++11-capable compiler, and SWIG 3.0 as system requirements. source
- Purpose
- Auto-PyTorch is an automated machine learning toolkit based on PyTorch that automates algorithm selection and hyperparameter tuning. source
- Resource controls
- Users can set a memory limit for estimators and a total wall-time limit for model search. source
- Search methods
- Its search process uses Bayesian optimization, meta-learning, SMAC, and Hyperband to explore pipeline configurations within a user-set budget. source
- Support and contribution
- The project invites bug reports, documentation improvements, and feature contributions through its GitHub issue tracker. source
- Support and contributions
- The project invites bug reports and documentation contributions through its GitHub issue tracker and recommends contacting developers by opening an issue before starting feature work. source
- Supported tasks
- The documentation describes tabular classification, tabular regression, and time-series forecasting tasks. source
- Supported tasks
- The toolkit supports tabular classification, tabular regression, and time series forecasting. source
- What it does
- Auto-PyTorch is an automated machine-learning toolkit based on PyTorch that helps users automate algorithm selection and hyperparameter tuning. source
Checked against the maker's pages on 3 Oct 2026.