Bound Frontier predictor training parallelism

runs/frontier-fidelity-envelope-v1/run_frontier_qwen30_exact_trace_surface.py

@@ -67,6 +67,7 @@ def parse_args() -> argparse.Namespace:
     )
     parser.add_argument("--allreduce-csv", type=Path)
     parser.add_argument("--timeout-seconds", type=float, default=1800.0)
+    parser.add_argument("--predictor-training-job-threads", type=int, default=1)
     parser.add_argument("--resume", action="store_true")
     parser.add_argument("--continue-on-failure", action="store_true")
     return parser.parse_args()
@@ -222,6 +223,8 @@ def score(path: Path, expected_shapes: list[tuple[int, int]]) -> dict[str, Any]:
 
 def main() -> None:
     args = parse_args()
+    if args.predictor_training_job_threads <= 0:
+        raise ValueError("predictor training job threads must be positive")
     for name in (
         "frontier_source",
         "replayserve_root",
@@ -298,6 +301,12 @@ def main() -> None:
                 run_id=f"qwen30_trace_{config.name}_{trace['label']}",
                 knobs=config_knobs,
             )
+            command.extend(
+                [
+                    "--random_forrest_execution_time_predictor_config_num_training_job_threads",
+                    str(args.predictor_training_job_threads),
+                ]
+            )
             command = BASE.configure_cc_command(
                 command,
                 backend=args.cc_backend,
@@ -471,6 +480,7 @@ def main() -> None:
             "tpot_slos_ms": TPOT_SLOS_MS,
             "primary_tpot_slo_ms": 150.0,
             "target_pass_rate": TARGET_PASS_RATE,
+            "predictor_training_job_threads": args.predictor_training_job_threads,
         },
         "frontier": {
             "source": str(args.frontier_source),