Upload build/publish_mix.py with huggingface_hub

build/publish_mix.py

@@ -133,6 +133,36 @@ def main():
     if not man.get("shards"):
         print("FATAL: refusing to publish a mix with zero shards", flush=True)
         sys.exit(5)
+    # §3.3 is a precondition of publishing, so it is enforced here rather than trusted to the caller: a
+    # contaminated mix cannot be shipped even by a script that forgot to check. The token-count equality is
+    # the part that catches the quiet failure -- an audit.json left behind by an earlier run of a *different*
+    # mix would otherwise certify bytes that were never scanned.
+    ap_path = os.path.join(a.root, "audit.json")
+    if not os.path.exists(ap_path):
+        print("FATAL: no audit.json in the mix root -- an unaudited mix is not publishable", flush=True)
+        sys.exit(10)
+    aud = json.load(open(ap_path))
+    problems = []
+    if not aud.get("AUDIT_PASSED"):
+        problems.append(f"AUDIT_PASSED={aud.get('AUDIT_PASSED')}")
+    if not aud.get("MEASURED"):
+        problems.append("MEASURED=false (some reference split was under-read or a loader failed)")
+    if not aud.get("COVERED_ALL_SHARDS"):
+        problems.append(f"shards missing from the audit: {aud.get('shards_missing')}")
+    if len(aud.get("tasks_covered") or []) != 8:
+        problems.append(f"only {len(aud.get('tasks_covered') or [])}/8 benchmark tasks covered")
+    if int(aud.get("mix_tokens_audited") or 0) != int(man.get("total_tokens") or -1):
+        problems.append(f"audit covered {aud.get('mix_tokens_audited')} tokens but the manifest holds "
+                        f"{man.get('total_tokens')}")
+    if problems:
+        print("FATAL: contamination gate refuses publication:", flush=True)
+        for p in problems:
+            print("  -", p, flush=True)
+        sys.exit(10)
+    print(f"audit gate: PASS k={aud.get('k')} overlap={aud.get('overlap_total')} "
+          f"(noise allowance {aud.get('noise_allowance')}), "
+          f"{len(aud.get('tasks_covered') or [])}/8 tasks, "
+          f"{aud.get('mix_documents_audited'):,} documents", flush=True)
     man["built_utc"] = __import__("time").strftime("%Y-%m-%dT%H:%M:%SZ", __import__("time").gmtime())
     with open(os.path.join(a.root, "manifest.json"), "w") as f:
         json.dump(man, f, indent=1)