gSampler

gSampler is a general and efficient GPU-based graph sampling technology specifically designed for graph learning. Using the "Extract-Compute-Select-Finalize" (ECSF) model for single-layer graph sampling, gSampler provides matrix-centric APIs that are user-friendly and intuitive.

To effectively accelerate the application of various sampling algorithms on different datasets, gSampler introduces data flow graphs to implement various optimizations, including computation optimizations (pre-processing, operator fusion, and common passes), data layout optimizations (automated format selection and graph compaction), and super-batching.

• Computation optimizations: eliminate redundant data movements and operators by optimizing on the data flow graph.
• Data layout optimizations: determine the most efficient data storage format to execute the program based on the sampling algorithm and dataset.
• Super-batching: batch the sampling tasks to improve GPU utilization without affecting training accuracy.

Compared to existing works, gSampler provides unprecedented programming support and performance improvement for graph sampling, especially for training graph neural networks (GNNs).

Install

Software Version

• Ubuntu 20.04
• g++/gcc v9.4.0
• CMake v3.16.3
• CUDA v11.6
• PyTorch v1.12.1
• DGL v0.9.1

Install gSampler

We use pip to manage our python environment.

1. Install PyTorch and DGL
2. Download gSampler systems

   git clone --recursive https://github.com/gsampler9/gSampler.git

3. Build and Install

   cd gsampler && mkdir build
   cd build && cmake ..
   # compile
   make -j12
   # install
   cd ../python && python3 setup.py install

Supported algorithms

• FastGCN
• LADIES
• AS-GCN
• GraphSAGE
• KGCN
• PASS
• ShadowGNN
• GraphSAINT
• RandomWalk
• Node2Vec

See example folder for details. More algorithms are being expanded.