Skip to content

v8.0-beta freeze: remaining HPC + external-artifact actions #5

Description

@jang1563

After PR #4 merged into `v3`, the v8.0-beta freeze has 3 input-side and 4 packaging-side blockers remaining. All require resources outside this local environment (HPC, HF, Zenodo).

Source of truth: `v8/provenance/input_freeze.json` (`release_blockers`) and `v8/release/v8_beta_artifact_manifest.json` (`release_blockers`) on the `v3` branch.

Input-side blockers (input_freeze.json)

1. HPC rebuild with L1000CDS2 pinned (input + code)

  • Switch `v8/intervene/lincs_query.py` and `v8/intervene/api_snapshot_manifest.py` `db-version` from `"latest"` to `"cpcd-gse70138-v1.0"` (most-recent named version per https://maayanlab.cloud/L1000CDS2/help)
  • Run `bash scripts/hpc_v8_beta_rebuild.sh --require-frozen` on HPC from a fresh clean checkout
  • Update affected run manifests in `v8/provenance/runs/` and `v8/intervene/evaluation/api_snapshot_manifest.json` payload SHA-256 with the actual rebuild output
  • Re-run `python3 scripts/validate_v8_provenance.py` until checksums all match

2. SpaceOmicsBench re-clone confirmation

  • Public re-clone: `git clone https://github.com/jang1563/SpaceOmicsBench.git && cd SpaceOmicsBench && git checkout 624048c5355e7381877357b8126e7d78606aa72a`
  • Verify the three CSV SHA-256 hashes in `v8/provenance/input_freeze.json` (`spaceomicsbench.v2_public` source) reproduce on the clean clone
  • If hashes mismatch, the local checkout has uncommitted modifications and the freeze record needs correction

3. HF + Zenodo artifact records

  • Resolve the final HF dataset bundle revision (HuggingFace commit/git SHA after upload) and record it in the artifact manifest
  • Mint or reserve a Zenodo DOI for the v8.0-beta release; record DOI URI in artifact manifest

Packaging-side blockers (v8_beta_artifact_manifest.json)

4. Upload HF dataset bundle

  • Build the v8 portion of the HF dataset bundle (BRIDGE / DECOMPOSE / INTERVENE compact evaluation outputs + provenance manifests)
  • Upload to `jang1563/genelab-benchmark` HF dataset and record the post-upload revision SHA in the manifest

5. Create Zenodo archive or reserve DOI

  • Either: enable GitHub-Zenodo integration so the next tag auto-mints a DOI
  • Or: manually upload the v8.0-beta source archive to Zenodo and record the DOI
  • A draft `.zenodo.json` on the main branch would automate metadata population (currently not present)

6. Archive raw external inputs outside Git with checksums

  • Stage raw OSDR / SpaceOmicsBench / SOMA / L1000CDS2 / Enrichr API payloads outside Git (object storage or Zenodo data artifact)
  • Record per-file SHA-256 in the artifact manifest under a new `raw_input_archive` section

7. Re-run hpc_v8_beta_rebuild.sh --require-frozen after frozen promotion

  • Once input_freeze.json status is promoted from `release_candidate` to `frozen`, run the rebuild gate one more time to confirm reproducibility
  • This is the final go/no-go gate for v8.0-beta tag

Suggested order

  1. Block Remove dangling refs to deleted DESIGN_DECISIONS / PLAN / PHASE1_RESULTS docs #2 (SpaceOmicsBench re-clone verify) — quick, no HPC needed; just a clean local clone
  2. Block Integrate v1-v7 work into main (v8 SpaceMed incubator excluded) #1 (HPC rebuild) — biggest item; includes the L1000CDS2 code change in the same commit
  3. Block v8: promote SpaceOmicsBench freeze to frozen after clean re-clone verify #6 (raw input archive) — can be staged in parallel with Integrate v1-v7 work into main (v8 SpaceMed incubator excluded) #1
  4. Block v8: close SpaceOmicsBench freeze blocker, promote L1000CDS2 to RC #4 (HF upload), Block v8.0-beta freeze: remaining HPC + external-artifact actions #5 (Zenodo DOI) — after Integrate v1-v7 work into main (v8 SpaceMed incubator excluded) #1 produces final artifacts
  5. Block v8/bridge: add exploratory mouse-tissue -> human i4_skin NES transfer #7 (re-run rebuild post-frozen) — final gate
  6. Block ci: add regression-tests workflow for PRs to main #3 (record HF/Zenodo IDs) — closes the manifest

Useful commands

```bash

0. Switch to v3 (v8 lives here)

git checkout v3

1. Local-side: pre-flight, verify status

python3 scripts/validate_v8_provenance.py

2. HPC: clean rebuild (assumes SPACEOMICS_ROOT is set)

bash scripts/hpc_v8_beta_rebuild.sh --require-frozen

3. Local-side: confirm SpaceOmicsBench re-clone hash

git clone https://github.com/jang1563/SpaceOmicsBench.git /tmp/sob-verify
cd /tmp/sob-verify && git checkout 624048c5355e7381877357b8126e7d78606aa72a
shasum -a 256 data/processed/gt_conserved_pathways_*.csv
```

Metadata

Metadata

Assignees

No one assigned

    Labels

    out-of-scope-v1-v7-releaseNot a blocker for the canonical v1-v7 public benchmark surfacev8-betav8 beta or translational-extension tracking item

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions