Setup Instructions on a GPU Machine

1. Install Conda

Follow instructions on the Conda user guide. For example, to install Miniconda on a linux machine:

wget https://repo.anaconda.com/miniconda/Miniconda3-py38_4.10.3-Linux-x86_64.sh
bash Miniconda3-py38_4.10.3-Linux-x86_64.sh

Or, to install Anaconda:

wget https://repo.anaconda.com/archive/Anaconda3-2021.05-Linux-x86_64.sh
bash Anaconda3-2021.05-Linux-x86_64.sh

SALIENT has been tested on Python 3.8.10.

It is highly recommended to create a new environment and do the subsequent steps therein. Example with a new environment called salient:

conda create -n salient python=3.8 -y
conda activate salient

2. Install PyTorch

Follow instructions on the PyTorch homepage. For example, to install on a linux machine with CUDA 11.3:

conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch

SALIENT has been tested on PyTorch 1.10.0.

3. Install PyTorch-Geometric (PyG)

Follow instructions on the PyG Github page. For example, with PyTorch >= 1.8.0, it suffices to do the following:

conda install pyg -c pyg -c conda-forge

SALIENT has been tested on PyG 2.0.2.

4. Install Latest PyTorch-Sparse

PyTorch-Sparse is usually installed together with PyG. We made a slight modification to support faster data transfer. See this pull request for detail. Currently, this change has been merged but not yet distributed. Hence, one should uninstall the prior version of PyTorch-Sparse and install the latest one from source:

pip uninstall torch_sparse
export FORCE_CUDA=1
pip install git+git://github.com/rusty1s/pytorch_sparse.git@master

Note: Compilation requires nvcc >= 11.

5. Install OGB

pip install ogb

SALIENT has been tested on OGB 1.3.2.

6. Install SALIENT's fast_sampler

Go to the folder fast_sampler and install:

cd fast_sampler
python setup.py install
cd ..

To check that it is properly installed, start python and type:

>>> import torch
>>> import fast_sampler
>>> help(fast_sampler)

One should see information of the package.

Note: Compilation requires a c++ compiler that supports c++17 (e.g., gcc >= 8).

7. Install Other Dependencies

conda install prettytable -c conda-forge

8. Try an Example

Congratulations! SALIENT has been installed. The script example_single_machine.sh under the folder examples contains examples to use SALIENT. Read it with care, edit as appropriate (e.g., set the correct SALIENT_ROOT and DATASET_ROOT), and run under this folder:

./example_single_machine.sh

It may take several minutes to run all examples therein.

8.1 Tips on Datasets

Create a folder somewhere to store the datasets. Pass the folder path to --dataset_root. The first time an OGB dataset is used, it will be automatically downloaded to that folder (which may take some time depending on size).

Alternatively, to pre-download an OGB dataset before trying the examples, start python and type:

>>> name = # type dataset name here, such as 'ogbn-arxiv'
>>> root = # type dataset root here, such as '/Users/username/dataset'
>>> from ogb.nodeproppred import PygNodePropPredDataset
>>> dataset = PygNodePropPredDataset(name=name, root=root)

Note: When an OGB dataset is used the first time, SALIENT will process it after downloading and will store the processed data under a processed subfolder of that dataset. Subsequent uses of SALIENT will directly load the processed data.

8.2 Tips on Command Line Arguments

To see all command-line arguments of SALIENT, set PYTHONPATH to be the root of SALIENT and type

python -m driver.main --help