Sampler with limit on nodes/edges in batch subgraph

[ayushnoori]

🚀 Feature

I'd like to implement a heterogeneous sampler that either sets an explicit upper bound on the number of nodes included in the message-flowing graph, or samples a fixed number of nodes in each iteration. This may be useful for DGL users who are learning on large heterogeneous graphs and need to explicitly limit the size of their subgraphs sampled in each batch (e.g., for memory purposes).

Motivation

Please see conversations with @jermainewang and @frozenbugs in the DGL Slack workspace. This issue is created to track the development of this feature. I will reference this issue in the relevant pull request.

@jermainewang @frozenbugs, please feel free to assign me to this feature.

[github-actions]

This issue has been automatically marked as stale due to lack of activity. It will be closed if no further activity occurs. Thank you