Archive 4 checkpoints (1-4)

paper-releases/pbisc-diffusion/updates/20261015T141849Z-rare-cell-7k-8k-panels/diffusion/dataset/server_1/model8/plans/rare_gene_context_expansion_20261015/selection_manifest.json

@@ -0,0 +1,138 @@
+{
+  "schema": "pbisc.rare_gene_expansion.v1",
+  "status": "review_only_panels_built",
+  "evidence_rows": 8774,
+  "evidence_requested_symbols": 4821,
+  "resolved_Ensembl_genes": 4163,
+  "curated_missing_from_final6k": 2091,
+  "curated_missing_train1m_positive": 2091,
+  "curated_missing_train1m_zero_detected": 0,
+  "core_raw_available_unique_genes": 218,
+  "core_missing_from6k": 37,
+  "protected_functional_sentinel_symbols": [
+    "AHR",
+    "AREG",
+    "CCR6",
+    "CSF2",
+    "IL13",
+    "IL17A",
+    "IL17F",
+    "IL17RB",
+    "IL1RL1",
+    "IL22",
+    "IL23R",
+    "IL5",
+    "NCR2",
+    "PPARG",
+    "RORC",
+    "TNFSF11",
+    "TNFSF4"
+  ],
+  "minimal6k_plus_core_count": 6037,
+  "minimal6k_plus_all_available_program_count": 8091,
+  "panels": [
+    {
+      "size": 7000,
+      "n_literature_additions": 1000,
+      "n_backfill_additions": 0,
+      "selected_genes_sha256": "b9ba04fb554cdec206a72fb6b00f2b3a7c49f038abebaa05473d87a71e265ec1",
+      "gene_positions_sha256": "f4bda0b94e2f5d6fcc7e44e34765a1b77b31dce85de49925bc95b3906ee91e10",
+      "manifest": {
+        "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_gene_context_expansion_20261015/panels/7k/selection_manifest.json",
+        "bytes": 6097,
+        "sha256": "c31f4adfcb2aa1692f2a54e967de1338276c4c98c7f5744fd97373cb476648d1"
+      }
+    },
+    {
+      "size": 8000,
+      "n_literature_additions": 2000,
+      "n_backfill_additions": 0,
+      "selected_genes_sha256": "62d961d83cf180dd925e080f696827f48074125cf2725d290cda21247d23c60f",
+      "gene_positions_sha256": "a6d8acb7ef27ae1c2733d50072e2d21cdb54b20a38910a5d10c4c8f9d72305d8",
+      "manifest": {
+        "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_gene_context_expansion_20261015/panels/8k/selection_manifest.json",
+        "bytes": 6097,
+        "sha256": "d9e46d3d6a0fbd7d7496c3f075a6c7c236eb5062926fd81f34bd071f2d79d8cb"
+      }
+    }
+  ],
+  "core_loss_7k": 0,
+  "core_loss_8k": 0,
+  "remaining_full_program_genes_7k": 1091,
+  "remaining_full_program_genes_8k": 91,
+  "decision_rule": "Both preserve core. Choose 7k for compact capability, or 8k for additional documented blood/state program coverage; neither panel size is proven necessary or sufficient for biological recovery. Full optional literature union is not a requirement.",
+  "no_active_configs_modified": true,
+  "training_launched": false,
+  "source_bindings": [
+    {
+      "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_gene_context_expansion_20261015/research/disease/evidence.csv",
+      "bytes": 1248197,
+      "sha256": "6a9d56309580cf74c4d7e74a5fe4f57b60ce558400e31a6a6af5df483f14317a"
+    },
+    {
+      "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_gene_context_expansion_20261015/research/ilc_dc/evidence.csv",
+      "bytes": 5966592,
+      "sha256": "4069ab07b995115741cc7649c36ba7e7366e9ab715f1bf5c3b49191cce7b8c61"
+    },
+    {
+      "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_gene_context_expansion_20261015/research/ilc_dc/protected_sentinels.csv",
+      "bytes": 5580,
+      "sha256": "76f7b9db90b6ec0319b0f68dc3e380dd852bb2dc2776b936b9149859453ed296"
+    },
+    {
+      "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_7_no_annotation_sh_hwang/data/e2_full_source/metadata/var.parquet",
+      "bytes": 755161,
+      "sha256": "64f4e30e17763bef37793e7229edb6caa0a7ffe3f3bfa5a08e1d4ca8ea7ea1dc"
+    },
+    {
+      "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_gene_context_expansion_20261015/reference/hgnc_complete_set_downloaded.tsv",
+      "bytes": 16982276,
+      "sha256": "2828618d23d315be76ea415f7ef0e951858716ebab07e2d584190f52e5178c1e"
+    },
+    {
+      "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_gene_context_expansion_20261015/reference/hgnc_download_receipt.json",
+      "bytes": 492,
+      "sha256": "666f522f8e15616e1fb747dd2d652949f60c1f7d226a8293fb26b4853f570a28"
+    },
+    {
+      "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/artifacts/panel_ladder_floor01_1m_20260923/manifest.json",
+      "bytes": 13488,
+      "sha256": "acbc8bc2cca5ecf3d3002e711b1d4a55b4069f17ae9666f0d64c197ca5cbc775"
+    },
+    {
+      "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/artifacts/panel_ladder_floor01_1m_20260923/full_gene_statistics.csv",
+      "bytes": 2919972,
+      "sha256": "745c8b5df2d1007f31e24e8391f15421885b78ca1c1026705ced4ae8d7591288"
+    },
+    {
+      "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/artifacts/panel_ladder_floor01_1m_20260923/ranked_gene_positions.npy",
+      "bytes": 160904,
+      "sha256": "4f6ce4fa492525cf06d68119a40c27d525ab6a36a4a99e09046684ae6edd767a"
+    },
+    {
+      "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/artifacts/panel_ladder_floor01_1m_20260923/tiers/6k/selected_genes.csv",
+      "bytes": 451366,
+      "sha256": "09248a3930e3fdefe1dbcd5d60cae4be81028526dcf097dd78169f5a415edfb6"
+    },
+    {
+      "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/artifacts/panel_ladder_floor01_1m_20260923/tiers/6k/gene_positions.npy",
+      "bytes": 48128,
+      "sha256": "653ffa45ac519000baa41086c52408009391d6e68cfb218bc6a50d34163fcb24"
+    },
+    {
+      "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_gene_context_expansion_20261015/audited_gene_universe.csv",
+      "bytes": 3759282,
+      "sha256": "4ba63a5eec12d308ee7f7656ef3a61c9295e7446a88efe92604734fe9fe0c0f6"
+    },
+    {
+      "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_gene_context_expansion_20261015/build_panels.py",
+      "bytes": 20137,
+      "sha256": "ec36872c10d3a560fa60ae8d42760cf70b596806453eedba7a69612a826302f7"
+    }
+  ],
+  "selection_uses_local_heldout_expression": false,
+  "selection_includes_benchmark_published_endpoint_markers": true,
+  "minimal6k_plus_core_and_protected_sentinels_count": 6044,
+  "core_and_protected_loss_7k": 0,
+  "core_and_protected_loss_8k": 0
+}

paper-releases/pbisc-diffusion/updates/20261015T141849Z-rare-cell-7k-8k-panels/diffusion/dataset/server_1/model8/plans/rare_gene_context_expansion_20261015/training_sampler_support.csv

@@ -0,0 +1,5 @@
+platform,recipes,unique_sample_keys,dataset_ids,donor_labels
+10x,4691,1992,19,1377
+Parse,2625,849,1,10
+ScaleBio,698,1,1,1
+Seq-Well,178,11,1,11

paper-releases/pbisc-diffusion/updates/20261015T141849Z-rare-cell-7k-8k-panels/diffusion/dataset/server_1/model8/plans/rare_gene_context_expansion_20261015/verify_panels.py

@@ -0,0 +1,79 @@
+"""Independent artifact invariants; no model imports or expression reads."""
+from __future__ import annotations
+
+import hashlib
+import json
+from pathlib import Path
+
+import numpy as np
+import pandas as pd
+
+
+def digest(path: Path) -> str:
+    value = hashlib.sha256()
+    with path.open("rb") as f:
+        for chunk in iter(lambda: f.read(1048576), b""):
+            value.update(chunk)
+    return value.hexdigest()
+
+
+def main() -> None:
+    out = Path(__file__).resolve().parent
+    model = out.parents[1]
+    old_dir = model / "artifacts/panel_ladder_floor01_1m_20260923/tiers/6k"
+    old = pd.read_csv(old_dir / "selected_genes.csv")
+    raw = pd.read_parquet(model.parent / "model_7_no_annotation_sh_hwang/data/e2_full_source/metadata/var.parquet")
+    summary = json.loads((out / "selection_manifest.json").read_text())
+    evidence = pd.read_csv(out / "evidence_resolved.csv").fillna("")
+    protected = set(evidence.loc[evidence.raw_available & evidence.selection_tier.eq(0), "gene_id"])
+    sources = []
+    for source in summary["source_bindings"]:
+        path = Path(source["path"])
+        assert path.stat().st_size == source["bytes"], str(path)
+        assert digest(path) == source["sha256"], str(path)
+        sources.append(source["path"])
+    assert digest(old_dir / "selected_genes.csv") == "09248a3930e3fdefe1dbcd5d60cae4be81028526dcf097dd78169f5a415edfb6"
+    assert digest(old_dir / "gene_positions.npy") == "653ffa45ac519000baa41086c52408009391d6e68cfb218bc6a50d34163fcb24"
+    last = set(old.gene_id)
+    verified = []
+    for size in [7000, 8000]:
+        panel_dir = out / "panels" / f"{size // 1000}k"
+        panel = pd.read_csv(panel_dir / "selected_genes.csv")
+        positions = np.load(panel_dir / "gene_positions.npy", allow_pickle=False)
+        mapping = pd.read_csv(panel_dir / "old_to_new.csv")
+        manifest = json.loads((panel_dir / "selection_manifest.json").read_text())
+        ids = set(panel.gene_id)
+        assert len(panel) == size == len(ids) == len(positions)
+        assert positions.dtype == np.int64 and np.all(np.diff(positions) > 0)
+        assert np.array_equal(panel.panel_position, positions)
+        assert np.array_equal(panel.tensor_index, np.arange(size))
+        assert np.array_equal(panel.gene_id, raw.iloc[positions].gene_id)
+        assert last < ids and set(old.gene_id) <= ids
+        assert protected <= ids
+        assert np.array_equal(mapping.old_tensor_index, np.arange(6000))
+        assert np.array_equal(mapping.gene_id, old.gene_id)
+        assert np.array_equal(panel.iloc[mapping.new_tensor_index].gene_id, old.gene_id)
+        new = panel[~panel.gene_id.isin(old.gene_id)]
+        assert len(new) == size - 6000
+        assert (new.measured_cells > 0).all()
+        assert (new.detected_cells > 0).all()
+        assert set(new.gene_id) <= set(evidence.loc[evidence.raw_available, "gene_id"])
+        assert new.selection_reason.eq("primary_literature_curated").all()
+        assert manifest["gene_positions_sha256"] == digest(panel_dir / "gene_positions.npy")
+        assert manifest["selected_genes_sha256"] == digest(panel_dir / "selected_genes.csv")
+        for source in manifest["source_bindings"]:
+            assert digest(Path(source["path"])) == source["sha256"]
+        last = ids
+        verified.append({"n_genes": size, "additions": len(new), "preserved_old_genes": 6000, "protected_core_and_sentinels_present": True, "coordinate_order_and_mapping": "pass", "all_added_genes_train1m_measured_and_detected": True, "selected_genes_sha256": digest(panel_dir / "selected_genes.csv"), "gene_positions_sha256": digest(panel_dir / "gene_positions.npy")})
+    # Check active model configurations still bind the immutable original panel.
+    configs = json.loads((out / "local_audit.json").read_text())["configs"]
+    for config in configs:
+        assert digest(Path(config["path"])) == config["sha256"]
+        assert config["same_final6k"]
+    receipt = {"status": "PASS", "checks": verified, "source_files_rehashed": len(sources), "active_6k_configs_unchanged": True, "no_cuda_imported": True, "no_training_or_inference": True, "biological_generation_improvement_tested": False, "verifier_sha256": digest(Path(__file__))}
+    (out / "VERIFIED.json").write_text(json.dumps(receipt, indent=2) + "\n")
+    print(json.dumps(receipt, indent=2))
+
+
+if __name__ == "__main__":
+    main()