Simple reference implementation of GraphSAGE.

Last update: Jan 6, 2023

Related tags

Overview

Reference PyTorch GraphSAGE Implementation

Author: William L. Hamilton

Basic reference PyTorch implementation of GraphSAGE. This reference implementation is not as fast as the TensorFlow version for large graphs, but the code is easier to read and it performs better (in terms of speed) on small-graph benchmarks. The code is also intended to be simpler, more extensible, and easier to work with than the TensorFlow version.

Currently, only supervised versions of GraphSAGE-mean and GraphSAGE-GCN are implemented.

Requirements

pytorch >0.2 is required.

Running examples

Execute python -m graphsage.model to run the Cora example. It assumes that CUDA is not being used, but modifying the run functions in model.py in the obvious way can change this. There is also a pubmed example (called via the run_pubmed function in model.py).

Comments

Can not run the sample until fix a little code in encoders.py

Hi William,

I clone the code and try run_cora. At first the interpreter raise the dimension problem until I fix the code in encoders.py(line 39) from:

neigh_feats = self.aggregator.forward(nodes, [self.adj_lists[node] for node in nodes],

to

neigh_feats = self.aggregator.forward(nodes, [self.adj_lists[int(node)] for node in nodes],

Then It works fine. I think it is a type issue. Hope it will be helpful for others.

opened by fs302 1
question about calling agg1 and enc1 twice

I wonder why do we need to call enc2 in self.features in both the encoder and aggregator? Why don't we just use the raw features as the agg2 already aggregates features from neighbors?

opened by sakhar 0
A question about the learning rate.

Hello!

I have a question about the learning rate. As is stated in the appendix of GraphSAGE, and are adopted by many other works, the learning rate is usually set to 1e-2, etc. Meanwhile, they usually normalize the input features.

However, in your work, the learning rate is set to 0.7, which is surprisingly high. You do not normalize the input features either. When I try to reset the learning rate to a common one and use the normalized features to train, I find that the model could only converge to a extremely bad performance.

This issue confuses me a lot. Could you help explain a bit?

opened by johannwyh 0
Potential typo - enc.num_sample[s]

In the model.py, whenever we are setting the number of samples for the encoders (enc1.num_samples), we should use enc1.num_sample rather than enc.num_samples

opened by kinkunchan 0
Can the code for inductive learning on the PPI dataset be added to the PyTorch implementation?

I'm following the PyTorch version of the GraphSage code and looks like the implementation is for a transductive setting (Cora and PubMed datasets). It would be really helpful if the implementation of the PPI dataset (inductive setting - graph level learning) can be added to the PyTorch implementation. In that way, one can train on a bunch of graphs and then test on a completely new set of graphs of different dimensions just like the PPI dataset.

opened by Sowmya-R-Krishnan 0
Is this for inductive learning?

Hello. It seems that GraphSage incorporates feature information from neighbors even if the neighbor belongs to test data while building the model. I think this is not allowed in inductive learning. Would you let me know if this code is for inductive learning?

opened by smayru 3
Is this pytorch version support batch-level learning?

Hi Dear Author,

Seems like the run functions are now targeting on transductive (node-level) classification. To implement the inductive (graph-level) with batched dataloader, we may need batch to store adjacencies and feature matrix, if original shape is (N, N) and (N, F) for them, now with batch they would change to (B, N, N) and (B, N, F), where B is batch-size, F is feature dim, N is node num.

Thus, to support batch-level calculation, is there any sections in Encoder and Aggregator need change?

opened by HarrialX 0

Simple reference implementation of GraphSAGE.

Related tags

Overview

Reference PyTorch GraphSAGE Implementation

Author: William L. Hamilton

Requirements

Running examples

Comments

Can not run the sample until fix a little code in encoders.py

question about calling agg1 and enc1 twice

A question about the learning rate.

Potential typo - enc.num_sample[s]

Can the code for inductive learning on the PPI dataset be added to the PyTorch implementation?

Is this for inductive learning?

Is this pytorch version support batch-level learning?

Owner

William L Hamilton

This repository contains numerical implementation for the paper Intertemporal Pricing under Reference Effects: Integrating Reference Effects and Consumer Heterogeneity.

A mini library for Policy Gradients with Parameter-based Exploration, with reference implementation of the ClipUp optimizer from NNAISENSE.

Fast, modular reference implementation of Instance Segmentation and Object Detection algorithms in PyTorch.

Reference implementation of code generation projects from Facebook AI Research. General toolkit to apply machine learning to code, from dataset creation to model training and evaluation. Comes with pretrained models.

This is the open-source reference implementation of the SIGGRAPH 2021 paper Intersection-free Rigid Body Dynamics.

[CVPR 2022] Official PyTorch Implementation for "Reference-based Video Super-Resolution Using Multi-Camera Video Triplets"

Intel® Nervana™ reference deep learning framework committed to best performance on all hardware

Image morphing without reference points by applying warp maps and optimizing over them.

Intel® Nervana™ reference deep learning framework committed to best performance on all hardware

MASA-SR: Matching Acceleration and Spatial Adaptation for Reference-Based Image Super-Resolution (CVPR2021)

Interpretation of T cell states using reference single-cell atlases

Code for C2-Matching (CVPR2021). Paper: Robust Reference-based Super-Resolution via C2-Matching.

A embed able annotation tool for end to end cross document co-reference

Reference code for the paper CAMS: Color-Aware Multi-Style Transfer.

Pip-package for trajectory benchmarking from "Be your own Benchmark: No-Reference Trajectory Metric on Registered Point Clouds", ECMR'21

EFENet: Reference-based Video Super-Resolution with Enhanced Flow Estimation

YouRefIt: Embodied Reference Understanding with Language and Gesture