Download scripts/lcb_grading.py from tts-sft/round4-sft-eval: direct link, hf CLI and curl.
- Browser
- Download file 5 kB
-
https://huggingface.co/datasets/tts-sft/round4-sft-eval/resolve/main/scripts/lcb_grading.py
- Command line
-
hf download hf://datasets/tts-sft/round4-sft-eval/scripts/lcb_grading.py
-
curl -L -o lcb_grading.py https://huggingface.co/datasets/tts-sft/round4-sft-eval/resolve/main/scripts/lcb_grading.py
5 kB
| #!/usr/bin/env python3 | |
| """P5 grading hardening: shared subprocess-exec wrapper with a PERSISTENT code-hash→verdict cache and | |
| best-of-N retry on timeout-category failures. | |
| Why: hidden grading uses per-test SIGALRM; candidates near the limit flip pass/fail with machine load | |
| (observed ±1–2 problems per pass on 126). Retrying ONLY timeout-category failures removes load flakes | |
| (a candidate that times out on every attempt stays failed); the cache makes repeated passes consistent | |
| and fast. Defaults keep the historical TLE (6 s/test) so numbers stay comparable; pass tle=10.0 for | |
| hardened reruns. The canonical score_se_subset.py pipeline is intentionally NOT modified (cross-node | |
| comparability) — use this in probes/ad-hoc graders. | |
| Usage: | |
| from lcb_grading import GradingCache, run_harness_cached | |
| cache = GradingCache("outputs/grading_cache/hidden.jsonl") | |
| v = run_harness_cached(HARNESS, code, tests_json, n_tests, cache=cache) # {"passed":bool,...} | |
| """ | |
| from __future__ import annotations | |
| import hashlib, json, os, subprocess, sys, tempfile, threading | |
| from pathlib import Path | |
| _TIMEOUT_MARKERS = ("timeout", "tle", "timeouterror", "timeoutexpired") | |
| def _is_timeout_verdict(v: dict | None) -> bool: | |
| if v is None: | |
| return True # parent-level harness timeout | |
| err = str(v.get("error", "")) + str(v.get("category", "")) | |
| return any(m in err.lower() for m in _TIMEOUT_MARKERS) | |
| class GradingCache: | |
| """Append-only JSONL cache: key -> verdict dict. Thread-safe; survives across passes/scripts.""" | |
| def __init__(self, path: str | os.PathLike): | |
| self.path = Path(path) | |
| self.path.parent.mkdir(parents=True, exist_ok=True) | |
| self._lock = threading.Lock() | |
| self._d: dict[str, dict] = {} | |
| if self.path.exists(): | |
| for line in self.path.open(): | |
| try: | |
| r = json.loads(line) | |
| self._d[r["k"]] = r["v"] | |
| except Exception: # noqa: BLE001 | |
| continue | |
| def get(self, k): | |
| return self._d.get(k) | |
| def put(self, k, v): | |
| with self._lock: | |
| if k in self._d: | |
| return | |
| self._d[k] = v | |
| with self.path.open("a") as f: | |
| f.write(json.dumps({"k": k, "v": v}, ensure_ascii=False) + "\n") | |
| def __len__(self): | |
| return len(self._d) | |
| def cache_key(harness, code: str, tests_json: str, tl: float) -> str: | |
| h = hashlib.md5() | |
| for part in (str(harness), code or "", tests_json, f"{tl}"): | |
| h.update(part.encode("utf-8", "ignore")); h.update(b"\x00") | |
| return h.hexdigest() | |
| def _run_once(harness, code: str, tests_json: str, n_tests: int, tl: float) -> dict | None: | |
| with tempfile.NamedTemporaryFile("w", suffix=".py", delete=False) as cf: | |
| cf.write(code); cp = cf.name | |
| with tempfile.NamedTemporaryFile("w", suffix=".json", delete=False) as tf: | |
| tf.write(tests_json); tp = tf.name | |
| try: | |
| p = subprocess.run([sys.executable, str(harness), cp, tp], capture_output=True, text=True, | |
| timeout=min(n_tests * tl + 20, 300)) | |
| out = p.stdout.strip().splitlines() | |
| r = json.loads(out[-1]) if out else None | |
| return r if isinstance(r, dict) else None # ignore non-dict last line (e.g. candidate stdout leaking an int) | |
| except Exception: # noqa: BLE001 (incl. parent TimeoutExpired) | |
| return None | |
| finally: | |
| Path(cp).unlink(missing_ok=True); Path(tp).unlink(missing_ok=True) | |
| def run_harness_cached(harness, code: str, tests_json: str, n_tests: int, *, | |
| tl: float = 6.0, retries_on_timeout: int = 1, | |
| cache: GradingCache | None = None) -> dict | None: | |
| """Run a candidate through a harness with caching + timeout-only retry. | |
| Retry rule: a non-timeout verdict (pass OR deterministic fail) is final on first attempt. | |
| A timeout-category failure is retried up to `retries_on_timeout` times; passing any attempt | |
| counts as pass (load flake), timing out every attempt stays a timeout failure. | |
| """ | |
| if not code: | |
| return {"passed": False, "error": "no_code", "category": "no_code"} | |
| k = cache_key(harness, code, tests_json, tl) if cache is not None else None | |
| if cache is not None: | |
| hit = cache.get(k) | |
| if hit is not None: | |
| return hit | |
| v = _run_once(harness, code, tests_json, n_tests, tl) | |
| attempts = 0 | |
| while _is_timeout_verdict(v) and not (v or {}).get("passed") and attempts < retries_on_timeout: | |
| attempts += 1 | |
| v2 = _run_once(harness, code, tests_json, n_tests, tl) | |
| if v2 is not None and (v2.get("passed") or not _is_timeout_verdict(v2)): | |
| v = v2 # flake resolved (pass, or a deterministic non-timeout failure) | |
| break | |
| v = v2 or v | |
| if v is None: | |
| v = {"passed": False, "error": "harness_timeout", "category": "timeout"} | |
| if cache is not None: | |
| cache.put(k, v) | |
| return v | |