Installation

Note

AutoGluon requires Python version 3.6 or 3.7. Linux is the only operating system fully supported for now (complete Mac OSX and Windows versions will be available soon). For troubleshooting the installation process, you can check the Installation FAQ.

Select your preferences below and run the corresponding install commands:

OS: Linux Mac

Linux.
Mac OSX.

Version: PIP Source

PIP Release.
Install AutoGluon from source.

Backend: CPU GPU

Build-in backend for CPU.
Required to run on Nvidia GPUs.

Command:

python3 -m pip install --upgrade "mxnet<2.0.0"
python3 -m pip install autogluon
# Here we assume CUDA 10.1 is installed.  You should change the number
# according to your own CUDA version (e.g. mxnet_cu100 for CUDA 10.0).
python3 -m pip install --upgrade "mxnet_cu101<2.0.0"
python3 -m pip install autogluon
python3 -m pip install --upgrade "mxnet<2.0.0"
git clone https://github.com/awslabs/autogluon
cd autogluon && python3 setup.py develop
# Here we assume CUDA 10.1 is installed.  You should change the number
# according to your own CUDA version (e.g. mxnet_cu102 for CUDA 10.2).
python3 -m pip install --upgrade "mxnet_cu101<2.0.0"
git clone https://github.com/awslabs/autogluon
cd autogluon && python3 setup.py develop

Note

If you don’t have them, please first install: XCode, Homebrew, LibOMP. Once you have Homebrew, LibOMP can be installed via:

brew install libomp
python3 -m pip install --upgrade "mxnet<2.0.0"
python3 -m pip install autogluon

Note

AutoGluon is not yet fully functional on Mac OSX. If you encounter MXNet system errors, please use Linux instead. However, you can currently use AutoGluon for less compute-intensive TabularPrediction tasks on your Mac laptop (but only with hyperparameter_tune = False).

Note

GPU usage is not yet supported on Mac OSX, please use Linux to utilize GPUs in AutoGluon.

Note

If you don’t have them, please first install: XCode, Homebrew, LibOMP. Once you have Homebrew, LibOMP can be installed via:

brew install libomp
python3 -m pip install --upgrade "mxnet<2.0.0"
git clone https://github.com/awslabs/autogluon
cd autogluon && python3 setup.py develop

Note

AutoGluon is not yet fully functional on Mac OSX. If you encounter MXNet system errors, please use Linux instead. However, you can currently use AutoGluon for less compute-intensive TabularPrediction tasks on your Mac laptop (but only with hyperparameter_tune = False).

Note

GPU usage is not yet supported on Mac OSX , please use Linux to utilize GPUs in AutoGluon.

Installation FAQ

  • Which version of MXNet does AutoGluon support?

    Currently, AutoGluon supports MXNet>=1.7.0. In order to ensure that you are installing mxnet larger than 1.7.0, you can use

    # For CPU
    python3 -m pip install "mxnet<2.0.0, >=1.7.0"
    
    # For GPU users, CUDA 101
    python3 -m pip install "mxnet_cu101<2.0.0, >=1.7.0"
    
  • I cannot install the package and it reports the error “XXX is not a supported wheel on this platform”.

    One possibility is that you are using an older version of pip. Try to upgrade your pip to a version later than “19.0.0”, e.g., use the following command:

    python3 -m pip install --upgrade pip --user
    
  • I see the error “ERROR: No matching distribution found for mxnet<2.0.0,>=1.7.0b20200713”.

    It might be due to the out-dated pip version. Try to upgrade the pip via:

    python3 -m pip install --upgrade pip --user
    
  • How can I install the customized mxnet (incubating) on SageMaker Notebook?

    You should choose the conda_python3 kernel and then install the MXNet via

    # For CPU users
    python3 -m pip install "mxnet<2.0.0"
    
    # For GPU users, CUDA 101
    python3 -m pip install "mxnet_cu101<2.0.0"
    
  • While running AutoGluon, I get error message “Check failed: e == cudaSuccess: CUDA: initialization error”.

    You may have the wrong version of MXNet installed for your CUDA version. Match the CUDA version carefully when following the installation instructions.

Next steps