Skip to content

Add US long-run bundle manifest#366

Merged
MaxGhenis merged 2 commits into
mainfrom
codex/crfb-longrun-bundle
May 17, 2026
Merged

Add US long-run bundle manifest#366
MaxGhenis merged 2 commits into
mainfrom
codex/crfb-longrun-bundle

Conversation

@MaxGhenis
Copy link
Copy Markdown
Contributor

Summary

Adds the CRFB US long-run managed dataset bundle to policyengine.py so downstream runs can resolve published 2026-2100 H5s through the normal managed dataset manifest instead of local policyengine-us / policyengine-us-data paths.

Key changes:

  • Updates the US release manifest to policyengine-us==1.691.12 and the crfb-longrun-20260517 policyengine-us-data release manifest.
  • Adds long-term dataset entries with H5 and metadata hashes.
  • Adds manifest resolution and US dataset helpers/tests for long-term managed datasets.
  • Pins the US extra to policyengine_core==3.26.1 with policyengine-us==1.691.12, matching the model package used to build the released long-run H5s.

Validation

  • uv run --with ruff ruff check src/policyengine/tax_benefit_models/us/datasets.py src/policyengine/provenance/manifest.py tests/test_us_long_term_datasets.py tests/test_release_manifests.py tests/test_models.py
  • uv run pytest tests/test_us_long_term_datasets.py tests/test_release_manifests.py tests/test_models.py -q (78 passed, 1 warning)

Notes

The long-run H5s are published at Hugging Face tag crfb-longrun-20260517; downstream CRFB sentinel runs have resolved and scored 2100 through this manifest path using the local worktree.

@MaxGhenis MaxGhenis marked this pull request as ready for review May 17, 2026 21:00
@MaxGhenis MaxGhenis merged commit 3cff439 into main May 17, 2026
11 checks passed
@MaxGhenis MaxGhenis deleted the codex/crfb-longrun-bundle branch May 17, 2026 21:00
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