exp: freeze raw KP traffic scaling control

1 files changed, 163 insertions, 0 deletions

diff --git a/experiments/analyze_kp_raw_traffic_scaling.py b/experiments/analyze_kp_raw_traffic_scaling.py
new file mode 100644
index 0000000..371406f
--- /dev/null
+++ b/experiments/analyze_kp_raw_traffic_scaling.py
@@ -0,0 +1,163 @@
+#!/usr/bin/env python3
+"""Audit the three-depth raw-KP mixed-traffic control."""
+import argparse
+import hashlib
+import json
+import math
+import os
+import sys
+
+
+ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+sys.path.insert(0, ROOT)
+from experiments.crossover_hardware import assert_hardware_report  # noqa: E402
+from experiments.kp_raw_traffic_scaling import (  # noqa: E402
+    DEPTHS,
+    RESULT_ROOT,
+    jobs,
+    registry_sha256,
+)
+
+
+def sha256(path):
+    digest = hashlib.sha256()
+    with open(path, "rb") as handle:
+        for chunk in iter(lambda: handle.read(1024 * 1024), b""):
+            digest.update(chunk)
+    return digest.hexdigest()
+
+
+def read_json(path):
+    with open(path, encoding="utf-8") as handle:
+        return json.load(handle)
+
+
+def finite(value):
+    return (
+        isinstance(value, (int, float))
+        and not isinstance(value, bool)
+        and math.isfinite(float(value))
+    )
+
+
+def audit_job(job, launch):
+    manifest_path = job["output"] + ".manifest.json"
+    if not os.path.isfile(manifest_path):
+        return {
+            "depth": job["depth"],
+            "architecture": job["architecture"],
+            "status": "missing_manifest",
+            "finite": False,
+            "final_validation_accuracy": None,
+        }
+    manifest = read_json(manifest_path)
+    assert manifest["source"] == launch["source"]
+    assert manifest["command"] == job["command"]
+    assert manifest["hardware_lock"]["physical_gpu_name"] == "NVIDIA RTX A6000"
+    assert_hardware_report(manifest["hardware_lock"], launch["hardware_policy"])
+    if manifest["status"] != "completed" or not manifest["output_exists"]:
+        return {
+            "depth": job["depth"],
+            "architecture": job["architecture"],
+            "status": manifest["status"],
+            "finite": False,
+            "final_validation_accuracy": None,
+            "driver_wall_seconds": manifest["driver_wall_seconds"],
+        }
+    assert sha256(job["output"]) == manifest["output_sha256"]
+    record = read_json(job["output"])
+    args = record["args"]
+    expected = {
+        "mode": "kp_traffic",
+        "traffic_rule": "raw",
+        "traffic_ratio": 4.0,
+        "traffic_seed": 5000,
+        "traffic_calibration_examples": 64,
+        "predictor_mode": "closed_form",
+        "predictor_warmup_steps": 1,
+        "predictor_every": 0,
+        "neutral_projection": 1,
+        "depth": job["depth"],
+        "width": 16,
+        "seed": 0,
+        "loader_seed": 0,
+        "split_seed": 2027,
+        "epochs": 200,
+        "eval_split": "validation",
+        "eval_every": 1,
+        "lr": 0.1,
+        "output_lr": 0.1,
+    }
+    for key, wanted in expected.items():
+        assert args[key] == wanted, (key, args[key], wanted)
+    assert record["provenance"] == {
+        "git_commit": launch["source"]["git_commit"],
+        "git_tracked_dirty": False,
+    }
+    assert record["split"]["train_examples"] == 45_000
+    assert record["split"]["validation_examples"] == 5_000
+    assert record["split"]["test_examples"] == 10_000
+    assert record["evaluation_protocol"]["test_evaluations"] == 0
+    assert record["counters"]["ordinary_examples"] == 9_000_000
+    assert record["counters"]["traffic_calibration_examples"] == 64
+    assert record["counters"]["predictor_update_examples"] == 64
+    assert record["counters"]["neutral_projection_examples"] == 9_000_000
+    assert record["counters"]["logical_batch_loss_queries"] == 0
+    final = record["final"]
+    complete = len(record["epochs"]) == 200
+    is_finite = bool(
+        complete
+        and final["finite"] is True
+        and finite(final["accuracy"])
+        and finite(final["loss"])
+    )
+    return {
+        "depth": job["depth"],
+        "architecture": job["architecture"],
+        "status": "completed" if complete else "incomplete_trajectory",
+        "finite": is_finite,
+        "final_validation_accuracy": (
+            float(final["accuracy"]) if finite(final["accuracy"]) else None
+        ),
+        "final_validation_loss": (
+            float(final["loss"]) if finite(final["loss"]) else None
+        ),
+        "first_nonfinite_step": record.get("first_nonfinite_step"),
+        "total_wall_seconds": record["timing"]["total_timed_wall_s"],
+        "peak_memory_allocated_bytes": record["hardware"][
+            "peak_memory_allocated_bytes"
+        ],
+    }
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--out", default=os.path.join(RESULT_ROOT, "r1_audit.json")
+    )
+    args = parser.parse_args()
+    launch_path = os.path.join(RESULT_ROOT, "r1_launch.json")
+    launch = read_json(launch_path)
+    registered = jobs()
+    assert launch["registry_sha256"] == registry_sha256(registered)
+    records = [audit_job(job, launch) for job in registered]
+    report = {
+        "audit_status": "passed",
+        "stage": "raw_kp_traffic_scaling_r1",
+        "complete_registry": True,
+        "num_expected_cells": len(DEPTHS),
+        "num_audited_cells": len(records),
+        "num_finite_cells": sum(record["finite"] for record in records),
+        "failure_retaining": True,
+        "test_policy": "none",
+        "records": records,
+    }
+    os.makedirs(os.path.dirname(os.path.abspath(args.out)), exist_ok=True)
+    with open(args.out, "w", encoding="utf-8") as handle:
+        json.dump(report, handle, indent=2, sort_keys=True)
+        handle.write("\n")
+    print(json.dumps(report, indent=2, sort_keys=True))
+
+
+if __name__ == "__main__":
+    main()