Skip to content
Snippets Groups Projects

Decision mask in arbter kernel output schema

Merged Egor Sobolev requested to merge feat/kernel-decision-in-output into master
2 unresolved threads

The downstream analysis need to know the decision of kernels.

@hammerd

Merge request reports

Approval is optional

Merged by David HammerDavid Hammer 6 months ago (Oct 2, 2024 1:16pm UTC)

Loading

Activity

Filter activity
  • Approvals
  • Assignees & reviewers
  • Comments (from bots)
  • Comments (from users)
  • Commits & branches
  • Edits
  • Labels
  • Lock status
  • Mentions
  • Merge request status
  • Tracking
    • Just to make sure I understand the goal: it's now enough to know the final decision mask, downstream also wants to know the masks resulting from each intermediary step of the selection plan? That seems fine, is presumably still small. Then I'd also update AdvancedFrameSelectionArbiter.on_matched_data to actually write this out.

    • Author Maintainer

      Correct. The the typical example, when the final decision is a mix of all hits and small random fraction of misses. Then the downstream processing is different for hits and misses. Arbiter kernels label data each for specific reason, which is a result itself.

    • Sure, that makes sense.

    • Please register or sign in to reply
397 397 except Exception as ex:
398 398 warn(f"Kernel for {prefix} failed: {ex}")
399 399 res[selection_name] = np.ones(num_frames, dtype=bool)
400 out_hash[f"{selection_name}.mask"] = list(map(bool, res[selection_name]))
  • merged

  • David Hammer mentioned in commit ec337400

    mentioned in commit ec337400

  • Please register or sign in to reply
    Loading