Repository navigation
perf: datacube shared_preloads runtime measurement - #47
Merged
Merged
Conversation
Self-contained profiler timing the cube likelihood with shared_preloads=False vs True (honest vmap timing, not a single jit on concrete params), reporting the realized speed-up. Includes the AUTOLENS_PROFILING_SMOKE short-circuit for CI. Provisional SMA ~4x; re-run at ALMA scale on an A100 for the authoritative figure. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds a focused runtime profiler for the datacube
shared_preloadsspeed-up: it times an identical-channel cube likelihood withshared_preloads=Falsevsshared_preloads=Trueand reports the ratio. Timing usesFitness._vmapover a batch rather than a single jit on concrete parameters (a single jit over fixed params can constant-fold the work and report a misleadingly fast time; the vmapped per-evaluation time is honest).Companion to the per-step decomposition in
likelihood_breakdown/datacube/delaunay.py. Correctness is asserted separately in the autolens_workspace_test parity script. Provisional SMA-scale result is ~4× per-likelihood; the authoritative figure is an ALMA-scale run on a quiet A100 (--instrument alma_high) — the ratio is the robust deliverable, absolute seconds are environment-bound. Includes the repo'sAUTOLENS_PROFILING_SMOKE=1lint short-circuit.Scripts Changed
likelihood_runtime/datacube/shared_preloads.py— new: shared-vs-unshared vmap runtime measurement with the CI smoke short-circuit.Upstream PR
Test Plan
AUTOLENS_PROFILING_SMOKE=1 python likelihood_runtime/datacube/shared_preloads.py(imports/setup) passes🤖 Generated with Claude Code