New in AnyLearning 0.26.3: large-table Tabular AI, Text AI and safe Hugging Face imports. Explore the guide →
Documentation
1. Installation

Installation

AnyLearning OSS runs on Windows, macOS and Linux. Everything it needs is in the download (Python, PyTorch and the pretrained weights included), so it has nothing to fetch after installation and works on a machine with no network.

Step 1: Download AnyLearning

  1. Visit the AnyLearning download page (opens in a new tab).
  2. Download the build for your operating system.
  3. Wait for the file to finish downloading. It is a few gigabytes, because the models ship with it.

Step 2: Install AnyLearning

For Windows

  1. For GPU training, install the NVIDIA driver (opens in a new tab) for your card first.
  2. Run the downloaded AnyLearning-Setup-*.exe.
  3. Follow the installer. It offers a Desktop shortcut and a Start menu entry.
  4. Launch AnyLearning from the shortcut.

For macOS

  1. Open the downloaded .dmg.
  2. Drag AnyLearning into your Applications folder.
  3. Eject the disk image.

macOS blocks applications that are not from the App Store, and the first launch may report the app as damaged. To clear that, open Terminal and run:

xattr -cr /Applications/AnyLearning.app
Damaged Warning

Then launch it again, and choose Open if macOS asks you to confirm.

For Linux

  1. For GPU training, install the NVIDIA driver from your distribution.
  2. Extract the archive:
tar -xzf AnyLearning-*-linux-x86_64.tar.gz
  1. Run it:
./AnyLearning/AnyLearning

The archive is self-contained and can live anywhere you can write to: no system packages are installed and nothing needs root.

Step 3: Start a project

Open AnyLearning and create a project. The open-source edition includes local training and does not require an account, license key, or online activation.

Where your data is kept

Projects, images, annotations, databases and trained models are written to ~/anylearning-data (C:\Users\<you>\anylearning-data on Windows). Installing a new version does not touch it, and uninstalling does not delete it.

System requirements

Windows10 or later, 64-bit
macOS11 or later, Apple Silicon or Intel
Linuxglibc 2.31 or later (Ubuntu 20.04+), 64-bit
Memory8 GB, 16 GB for training larger models
Disk10 GB for the application, plus room for your datasets
GPU (optional)NVIDIA with CUDA, on Windows and Linux

Training works without a GPU: it is slower, not unavailable. On macOS, training runs on the CPU: Apple's GPU backend is not used yet.

Troubleshooting

  • macOS says the app is damaged: run the xattr command in Step 2.
  • Blank window on first launch: close and reopen the app. The frontend is unpacked on the first run, which takes a few seconds.
  • Training ends immediately with "No model found in training output": your batch size is larger than the number of images in a split. Lower it, or add images.

For anything else, visit our contact page (opens in a new tab) or email support@nrl.ai.

Introduction to AnyLearning2. Image Classification