Spaces:
Running
Running
File size: 13,554 Bytes
df97e68 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 | """
Pre-train checklist + GO/NO-GO gate for Gov Workflow RL Phase 1.
This script validates the local training stack without running training.
Use it before starting Phase 1 retraining.
Usage:
python scripts/pretrain_go_nogo.py
python scripts/pretrain_go_nogo.py --run-tests
"""
from __future__ import annotations
import argparse
import importlib
import json
import subprocess
import sys
from dataclasses import asdict, dataclass
from pathlib import Path
from typing import Callable
ROOT = Path(__file__).resolve().parents[1]
if str(ROOT) not in sys.path:
sys.path.insert(0, str(ROOT))
PHASE1_TASK = "district_backlog_easy"
EXPECTED_OBS_DIM = 84
EXPECTED_ACTIONS = 28
@dataclass
class CheckResult:
name: str
status: str # PASS | WARN | FAIL
detail: str
def _run_cmd(cmd: list[str], cwd: Path | None = None) -> tuple[int, str, str]:
proc = subprocess.run(
cmd,
cwd=str(cwd or ROOT),
capture_output=True,
text=True,
)
return proc.returncode, proc.stdout, proc.stderr
def check_required_files() -> CheckResult:
required = [
"rl/train_ppo.py",
"rl/train_recurrent.py",
"rl/gov_workflow_env.py",
"rl/feature_builder.py",
"rl/action_mask.py",
"rl/callbacks.py",
"rl/curriculum.py",
"rl/cost_tracker.py",
"rl/evaluate.py",
"rl/eval_grader.py",
"rl/plot_training.py",
"rl/configs/ppo_easy.yaml",
"app/env.py",
"app/models.py",
"app/tasks.py",
"app/reward.py",
"app/graders.py",
]
missing = [p for p in required if not (ROOT / p).exists()]
if missing:
return CheckResult(
name="required_files",
status="FAIL",
detail="Missing files: " + ", ".join(missing),
)
return CheckResult(
name="required_files",
status="PASS",
detail=f"{len(required)} required files present",
)
def check_python_imports() -> CheckResult:
modules = [
"yaml",
"numpy",
"gymnasium",
"torch",
"stable_baselines3",
"sb3_contrib",
"tensorboard",
"rl.train_ppo",
"rl.train_recurrent",
"rl.gov_workflow_env",
"rl.feature_builder",
"rl.action_mask",
"rl.callbacks",
"rl.evaluate",
"rl.eval_grader",
"app.env",
"app.tasks",
"app.graders",
]
failed: list[str] = []
for mod in modules:
try:
importlib.import_module(mod)
except Exception:
failed.append(mod)
if failed:
return CheckResult(
name="python_imports",
status="FAIL",
detail="Import failures: " + ", ".join(failed),
)
return CheckResult(
name="python_imports",
status="PASS",
detail=f"{len(modules)} modules import cleanly",
)
def check_compile() -> CheckResult:
targets = [
"rl/train_ppo.py",
"rl/train_recurrent.py",
"rl/gov_workflow_env.py",
"rl/feature_builder.py",
"rl/action_mask.py",
"rl/callbacks.py",
"rl/evaluate.py",
"rl/eval_grader.py",
"app/env.py",
"app/reward.py",
"app/graders.py",
"app/tasks.py",
]
cmd = [sys.executable, "-m", "py_compile", *targets]
rc, _out, err = _run_cmd(cmd, ROOT)
if rc != 0:
return CheckResult(
name="py_compile",
status="FAIL",
detail=err.strip() or "py_compile failed",
)
return CheckResult(
name="py_compile",
status="PASS",
detail=f"{len(targets)} files compiled successfully",
)
def check_env_contract() -> CheckResult:
try:
from rl.gov_workflow_env import GovWorkflowGymEnv
env = GovWorkflowGymEnv(task_id=PHASE1_TASK, seed=42)
obs, info = env.reset(seed=42)
masks = env.action_masks()
_obs2, reward, terminated, truncated, step_info = env.step(18)
problems: list[str] = []
if tuple(obs.shape) != (EXPECTED_OBS_DIM,):
problems.append(f"obs shape={tuple(obs.shape)} expected={(EXPECTED_OBS_DIM,)}")
if int(env.action_space.n) != EXPECTED_ACTIONS:
problems.append(f"action_space={env.action_space.n} expected={EXPECTED_ACTIONS}")
if len(masks) != EXPECTED_ACTIONS:
problems.append(f"mask_len={len(masks)} expected={EXPECTED_ACTIONS}")
if int(sum(bool(x) for x in masks)) <= 0:
problems.append("all actions masked")
if not isinstance(info, dict):
problems.append("reset info is not dict")
if not isinstance(step_info, dict):
problems.append("step info is not dict")
if not isinstance(float(reward), float):
problems.append("reward not float-castable")
if not isinstance(bool(terminated), bool) or not isinstance(bool(truncated), bool):
problems.append("terminated/truncated invalid type")
if problems:
return CheckResult(
name="gym_env_contract",
status="FAIL",
detail="; ".join(problems),
)
return CheckResult(
name="gym_env_contract",
status="PASS",
detail=f"obs={obs.shape}, action_n={env.action_space.n}, valid_masks={int(sum(masks))}",
)
except Exception as exc:
return CheckResult(
name="gym_env_contract",
status="FAIL",
detail=f"{type(exc).__name__}: {exc}",
)
def check_output_paths() -> CheckResult:
needed_dirs = [
ROOT / "results",
ROOT / "results" / "best_model",
ROOT / "results" / "runs",
ROOT / "results" / "eval_logs",
ROOT / "logs",
]
try:
for d in needed_dirs:
d.mkdir(parents=True, exist_ok=True)
probe = d / ".write_probe.tmp"
probe.write_text("ok", encoding="utf-8")
probe.unlink(missing_ok=True)
except Exception as exc:
return CheckResult(
name="output_paths",
status="FAIL",
detail=f"{type(exc).__name__}: {exc}",
)
return CheckResult(
name="output_paths",
status="PASS",
detail="results/ and logs/ are writable",
)
def check_train_cli() -> CheckResult:
cmd = [sys.executable, "-m", "rl.train_ppo", "--help"]
rc, out, err = _run_cmd(cmd, ROOT)
if rc != 0:
return CheckResult(
name="train_cli",
status="FAIL",
detail=err.strip() or "train_ppo --help failed",
)
needed_flags = [
"--phase",
"--timesteps",
"--n_envs",
"--task",
"--phase1-eval-freq",
"--phase1-n-eval-episodes",
"--phase1-disable-eval-callback",
"--phase1-grader-eval-freq-multiplier",
]
missing = [f for f in needed_flags if f not in out]
if missing:
return CheckResult(
name="train_cli",
status="WARN",
detail="Missing expected flags in help output: " + ", ".join(missing),
)
return CheckResult(
name="train_cli",
status="PASS",
detail="train_ppo CLI flags detected",
)
def check_config() -> CheckResult:
try:
import yaml
cfg_path = ROOT / "rl" / "configs" / "ppo_easy.yaml"
cfg = yaml.safe_load(cfg_path.read_text(encoding="utf-8-sig")) or {}
hp = cfg.get("hyperparameters", {})
tr = cfg.get("training", {})
required_fields = [
("hyperparameters", "learning_rate"),
("hyperparameters", "n_steps"),
("hyperparameters", "batch_size"),
("training", "n_envs"),
("training", "seed"),
("training", "eval_freq"),
("training", "n_eval_episodes"),
]
missing = []
for section, key in required_fields:
parent = hp if section == "hyperparameters" else tr
if key not in parent:
missing.append(f"{section}.{key}")
if missing:
return CheckResult(
name="ppo_easy_config",
status="FAIL",
detail="Missing config fields: " + ", ".join(missing),
)
warnings = []
if int(tr.get("eval_freq", 0)) < 2048:
warnings.append("eval_freq is very low; may cause frequent pauses")
if int(tr.get("n_eval_episodes", 0)) > 5:
warnings.append("n_eval_episodes is high; callback cost may increase")
if warnings:
return CheckResult(
name="ppo_easy_config",
status="WARN",
detail="; ".join(warnings),
)
return CheckResult(
name="ppo_easy_config",
status="PASS",
detail="Phase 1 config fields are present and reasonable",
)
except Exception as exc:
return CheckResult(
name="ppo_easy_config",
status="FAIL",
detail=f"{type(exc).__name__}: {exc}",
)
def check_torch_device() -> CheckResult:
try:
import torch
if torch.cuda.is_available():
return CheckResult(
name="torch_device",
status="PASS",
detail=f"CUDA available ({torch.cuda.get_device_name(0)})",
)
return CheckResult(
name="torch_device",
status="WARN",
detail="CUDA not available; CPU training is expected",
)
except Exception as exc:
return CheckResult(
name="torch_device",
status="WARN",
detail=f"torch device check skipped: {type(exc).__name__}: {exc}",
)
def run_targeted_tests() -> CheckResult:
test_cmd = [
sys.executable,
"-m",
"pytest",
"tests/test_env.py",
"tests/test_gym_wrapper.py",
"tests/test_gym_wrapper_integration.py",
"tests/test_feature_builder.py",
"tests/test_action_mask.py",
"tests/test_curriculum.py",
"tests/test_rl_evaluate.py",
"-q",
"--tb=short",
]
rc, out, err = _run_cmd(test_cmd, ROOT)
if rc != 0:
return CheckResult(
name="targeted_tests",
status="FAIL",
detail=(out + "\n" + err).strip()[-1200:],
)
return CheckResult(
name="targeted_tests",
status="PASS",
detail=out.strip().splitlines()[-1] if out.strip() else "targeted tests passed",
)
def _print_results(results: list[CheckResult]) -> None:
print("\n=== Pre-Train Checklist Results ===")
for r in results:
print(f"[{r.status}] {r.name}: {r.detail}")
fail_count = sum(1 for r in results if r.status == "FAIL")
warn_count = sum(1 for r in results if r.status == "WARN")
print("\n=== Gate Decision ===")
if fail_count > 0:
print(f"NO-GO: {fail_count} failing check(s). Resolve failures before training.")
else:
print(
f"GO: no failing checks. "
f"{warn_count} warning(s) can be reviewed but do not block training."
)
def _print_next_commands(args: argparse.Namespace) -> None:
print("\n=== Recommended Phase 1 Commands (Manual) ===")
print(
"python -m rl.train_ppo "
f"--phase 1 --task {PHASE1_TASK} "
f"--timesteps {args.timesteps} --n_envs {args.n_envs} --seed {args.seed} "
"--phase1-no-progress-bar "
"--phase1-eval-freq 16384 "
"--phase1-n-eval-episodes 2 "
"--phase1-grader-eval-freq-multiplier 4"
)
print(
"python rl/eval_grader.py "
"--model results/best_model/phase1_final "
f"--task {PHASE1_TASK} --episodes 20 --seed {args.seed}"
)
print(
"python rl/plot_training.py "
f"--task {PHASE1_TASK} --phase 1"
)
def main() -> int:
parser = argparse.ArgumentParser(description="Pre-train checklist + GO/NO-GO gate")
parser.add_argument("--run-tests", action="store_true", help="Run targeted RL tests")
parser.add_argument("--timesteps", type=int, default=300000)
parser.add_argument("--n-envs", "--n_envs", dest="n_envs", type=int, default=4)
parser.add_argument("--seed", type=int, default=42)
parser.add_argument("--json-out", default=None, help="Optional path to write JSON report")
args = parser.parse_args()
checks: list[Callable[[], CheckResult]] = [
check_required_files,
check_python_imports,
check_compile,
check_train_cli,
check_config,
check_env_contract,
check_output_paths,
check_torch_device,
]
if args.run_tests:
checks.append(run_targeted_tests)
results = [fn() for fn in checks]
_print_results(results)
_print_next_commands(args)
if args.json_out:
out_path = Path(args.json_out)
out_path.parent.mkdir(parents=True, exist_ok=True)
payload = {
"go_no_go": "NO-GO" if any(r.status == "FAIL" for r in results) else "GO",
"results": [asdict(r) for r in results],
}
out_path.write_text(json.dumps(payload, indent=2), encoding="utf-8")
print(f"\nJSON report written to: {out_path}")
return 2 if any(r.status == "FAIL" for r in results) else 0
if __name__ == "__main__":
raise SystemExit(main())
|