pyg-team/pytorch_geometric

Improve `NNConv` runtimes

Open

#602 opened on Aug 6, 2019

View on GitHub
 (8 comments) (0 reactions) (0 assignees)Python (3,514 forks)batch import
2 - Priority P2benchmarkfeaturehelp wantednn

Repository metrics

Stars
 (19,985 stars)
PR merge metrics
 (Avg merge 35d 1h) (14 merged PRs in 30d)

Description

🐛 Bug

I have benchmarked EEC vs https://github.com/mys007/ecc implementation. The code is 3-5 times slower in pytorch_geometric.

I think it is due to the entire pseudo tensor which is given to every message function. Maybe, only weight_j should be given there.

To Reproduce

Steps to reproduce the behavior:

Expected behavior

Environment

  • OS:
  • Python version:
  • PyTorch version:
  • CUDA/cuDNN version:
  • GCC version:
  • Any other relevant information:

Additional context

Contributor guide

Research direction
Profile the NNConv forward pass to identify bottlenecks, particularly in message computation where the full pseudo tensor is used.
Tech stack
pythonpytorch
Domain
machine learning
Issue type
Bug
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
2
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
1-3 hours
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
Active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Mostly clear
Prerequisites
PythonPyTorchgraph neural networks
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
50