Installation#

System Requirements#

Mammoth Lite requires:

Python: 3.10 or higher
Operating System: Linux, macOS, or Windows
GPU: Optional but recommended for faster training

Dependencies#

Mammoth Lite has minimal dependencies to keep the framework lightweight:

Core Dependencies:

torch - PyTorch for deep learning
torchvision - Computer vision utilities
numpy - Numerical computing
tqdm - Progress bars
gitpython - Git repository handling
six - Python 2/3 compatibility

Development Dependencies:

pytest - Testing framework
ipywidgets - Jupyter notebook widgets
mypy - Type checking

Installation Methods#

Method 1: Development Installation (Recommended)#

For development or if you want to modify the code:

# Clone the repository
git clone https://github.com/aimagelab/mammoth-lite.git
cd mammoth-lite

# Install in development mode
pip install -e .

This installs Mammoth Lite in “editable” mode, so changes to the source code are immediately available.

Method 2: Direct Installation#

If you just want to use Mammoth Lite without modifying it:

pip install mammoth-lite

Method 3: From Source#

# Download and extract the source
wget https://github.com/aimagelab/mammoth-lite/archive/main.zip
unzip main.zip
cd mammoth-lite-main

# Install
pip install .

Using with UV (Recommended)#

If you’re using the uv package manager:

# Install dependencies
uv sync

# Run with uv
uv run python mammoth_lite/main.py --model sgd --dataset seq-cifar10

GPU Support#

Mammoth Lite automatically uses GPU acceleration if available. To install PyTorch with CUDA support:

# For CUDA 11.8 (adjust version as needed)
pip install torch torchvision --index-url https://download.pytorch.org/whl/cu118

Verification#

To verify your installation works correctly:

# Test basic import
import mammoth_lite

# Check available models and datasets
from mammoth_lite import get_model_names, get_dataset_names

print("Available models:", get_model_names())
print("Available datasets:", get_dataset_names())

You should see output listing the available models and datasets.

Running a Quick Test#

Test that everything works with a quick training run:

cd mammoth_lite
python main.py --model sgd --dataset seq-cifar10 --n_epochs 1

This should start training an SGD model on Sequential CIFAR-10 for 1 epoch.

Next Steps#

Once you have Mammoth Lite installed, continue to the Quickstart Guide guide to run your first continual learning experiment!