Certify US Populace default: Build I 57k (rmloss100)#463
Merged
Conversation
…35, misc in-band, loss 0.030833 (populace#359) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The certify_data_release run updated the bundle manifest, pyproject, and changelog fragment but did not regenerate the derived US TRO sidecar, leaving it pinned to the previous certification (populace-us-2024-sparse-l0-refit-57k, policyengine-us 1.752.2). Regenerate src/policyengine/data/bundle/us.trace.tro.jsonld via the generate_trace_tros code path (US only, so the unchanged UK sidecar and its private-repo data_release_manifest artifact are left intact). The sidecar now binds: - bundle_manifest sha256 dfa2e5d8659f43cdb5d8b67585e88bfd08009100ae96e09d4f7d0c1fc22d738c - data_release_manifest sha256 c2bbe50a75279a83f741ec2267134702ee0e1de2e7babb74bd72ef148d25aa4b - dataset sha256 5e2a10b97f1fa740fbc1cc1be3450a19b2089c56de59d3754db13ec0dd048b88 - policyengine-us 1.764.6 wheel Fixes the stale bundle_manifest sha256 that failed TestVendoredSidecarBinding::test__given_vendored_bundle_manifest__then_tro_sidecar_binds_it. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Repoint the test constants and model surface snapshot that pinned the previous US certification to the certified Build I 57k release (populace-us-2024-buildi-sparse-rmloss100-6e8e929-20260709T034135Z, policyengine-us 1.764.6): - test_release_manifests.py: US_MODEL_VERSION, US_BUILT_WITH_MODEL_VERSION, US_DATA_RELEASE_ID (the derived URI/path/revision constants and the offline-fallback certification assertions follow from these). - test_models.py: US model_package.version and default_dataset_uri. - test_us_regions.py: national region dataset_path. - us_model_surface.json: regenerated via PE_UPDATE_SNAPSHOTS=1; the 1.752.2 -> 1.764.6 model bump shifts the bucketed counts (parameters 969 -> 978, variables 56 -> 57). Co-Authored-By: Claude Fable 5 <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
1 participant
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.
Certifies the Build I 57k artifact as the US default, replacing populace-us-2024-sparse-l0-refit-57k-71a0887 (July 1).
Release:
populace-us-2024-buildi-sparse-rmloss100-6e8e929-20260709T034135Z(latest.json already flipped via the tracked publish)Dataset sha256:
5e2a10b9…· 57,240 households · pe-us 1.764.6 · generated byscripts/certify_data_release.pyWhy promote (promotion scorecard — populace#359 verdict, all figures verbatim from artifacts)
First sparse artifact to pass every gate including export-mass parity (SOI identification from populace#299 Build H + selection-support fix from #359; only estate_income + non_sch_d_capital_gains carry band-math-justified reviewed exclusions). Selection = 99.825% overlap with the frozen 57,240 (100-record misc-support swap).
Known scope caveats carried from #299/#359: populace#340 never-persisted input families (unchanged from the incumbent), and the optional reform-validation tail trips a pre-existing pe-us MD UNKNOWN-county fragility (policyengine-us#8975, populace#367) — the artifact itself is geographically clean (0 UNKNOWN counties).
🤖 Generated with Claude Code