pyg-team/pytorch_geometric

List node features can not be sliced with subgraph

Open

#6,162 opened on Dec 7, 2022

View on GitHub
 (13 comments) (0 reactions) (0 assignees)Python (3,514 forks)batch import
0 - Priority P0featurehelp wanted

Repository metrics

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

Description

How to pass subgraph Data as a node feature into main graph Data? test_data=Data(edge_index=edge_index,pos=torch.tensor(np.array(pos)), graph=subgraphList) subgraphList is as follows:

 Data(x=[1, 3], edge_index=[2, 0], pos=[1, 2]),
 ...
 Data(x=[10, 3], edge_index=[2, 18])]```
The issue is that when I tried to mask some nodes:
`mask_test_data=test_data.subgraph(mask)`
mask_test_data.graph is the same as test_data.graph, while the size of mask_test_data.pos is reduced from test_data.pos.

After testing, I found the issue is caused by the type of list. May I know whether there is another type of container in which I can store the subgraph Data, and it can auto-align with  `test_data.subgraph(mask)`?

[Slack Message](https://torchgeometricco.slack.com/archives/C01DN0B3B1N/p1670346485240969?thread_ts=1670346485.240969&cid=C01DN0B3B1N)

Contributor guide

Research direction
Inspect the subgraph() method in PyG's Data class to understand how it copies attributes. The issue is that list attributes are not being sliced like tensor attributes. Look into the cat dim and inc methods for custom data types. Consider using a custom container or modifying the attribute handling to support slicing.
Tech stack
pythonpytorch
Domain
None
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
PythonPyTorchPyG
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
60