Skip to content

Conversation

brettviren
Copy link
Member

This PR collects changes needed to produce live+dead tiling, live+dead depo blob fill and live charge solving data for GNN training.

Things this PR provides (needs to provide)

  • ClusterFanin
  • a component to form blob-blob edges between blobs in an existing cluster (specifically live-dead connections)
  • job configuration

This PR is developed top of #193 so that PR needs merging first.

CHANGE from using FrameSplitter to FrameFanout.  This is done because
apparently TbbFlow does not yet have a wrapper for FrameSplitter.
This change SHOULD be innocuous.

ADD a new depo set version of the older per-depo simulation subgraph
construction.  This exposes a new `signal_sets` which is like `signal`
but the edges transfer depo sets instead of single depos.  This is
done to avoid the "Pgrapher slowdown" problem.  We should probably
make this new `signal_sets` be default by naming it `signal`.
@brettviren
Copy link
Member Author

Comment to mark that last commit touches MicroBooNE configuration. It should not change existing jobs but we had best apply a careful test. Let's bake that test into a BATS file.

I believe we should make new signal_sets become default signal for MicroBooNE simulation jobs so the graph passes depo sets instead of single depos to avoid Pgrapher slowdown problem.

@brettviren brettviren marked this pull request as ready for review March 20, 2023 12:40
@brettviren brettviren changed the title Produce GNN training samples Prepare WCT to make GNN training samples Sep 12, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant