File size: 5,859 Bytes
80d8c84
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
#!/usr/bin/env python3
"""Run a single ReplicaLab episode locally and dump logs.

OBS 07 — Quick local smoke-test script.  Resets the environment with a
given seed/scenario/difficulty, runs a baseline propose→accept sequence,
and writes the episode replay JSON + reward CSV/JSONL to the default
output directories.

Usage:
    python -m scripts.run_episode
    python -m scripts.run_episode --seed 42 --scenario ml_benchmark --difficulty hard
    python -m scripts.run_episode --max-rounds 3
"""

from __future__ import annotations

import argparse
import sys
from pathlib import Path

# Ensure project root is importable when run as a script
_PROJECT_ROOT = Path(__file__).resolve().parents[1]
if str(_PROJECT_ROOT) not in sys.path:
    sys.path.insert(0, str(_PROJECT_ROOT))

from replicalab.config import DEFAULT_DIFFICULTY, DEFAULT_SCENARIO_TEMPLATE
from replicalab.env.replicalab_env import ReplicaLabEnv
from replicalab.models import ScientistAction
from replicalab.scenarios import generate_scenario
from replicalab.utils.logging import log_episode_reward, write_episode_log


def _build_propose_action(env: ReplicaLabEnv, seed: int, scenario: str, difficulty: str) -> ScientistAction:
    """Build a baseline propose_protocol action from the scenario pack."""
    pack = generate_scenario(seed=seed, template=scenario, difficulty=difficulty)
    lab = pack.lab_manager_observation
    spec = pack.hidden_reference_spec
    return ScientistAction(
        action_type="propose_protocol",
        sample_size=10,
        controls=["baseline", "ablation"],
        technique=spec.summary[:60] if spec.summary else "replication_plan",
        duration_days=max(1, min(2, lab.time_limit_days)),
        required_equipment=list(lab.equipment_available[:1]) if lab.equipment_available else [],
        required_reagents=list(lab.reagents_in_stock[:1]) if lab.reagents_in_stock else [],
        questions=[],
        rationale=(
            f"Plan addresses: {', '.join(spec.required_elements[:2])}. "
            f"Target metric: {spec.target_metric}. "
            f"Target value: {spec.target_value}. "
            "Stay within budget and schedule."
        ),
    )


def _build_accept_action() -> ScientistAction:
    return ScientistAction(
        action_type="accept",
        sample_size=0,
        controls=[],
        technique="",
        duration_days=0,
        required_equipment=[],
        required_reagents=[],
        questions=[],
        rationale="",
    )


def run_episode(
    seed: int = 0,
    scenario: str = DEFAULT_SCENARIO_TEMPLATE,
    difficulty: str = DEFAULT_DIFFICULTY,
    max_rounds: int | None = None,
) -> None:
    """Run one episode and persist outputs."""
    env = ReplicaLabEnv()
    obs = env.reset(seed=seed, scenario=scenario, difficulty=difficulty)
    episode_id = env.episode_id()

    print(f"Episode {episode_id} | seed={seed} scenario={scenario} difficulty={difficulty}")
    print(f"  Paper: {obs.scientist.paper_title}")

    propose_action = _build_propose_action(env, seed, scenario, difficulty)
    total_steps = 0
    invalid_count = 0

    # Step 1: propose
    result = env.step(propose_action)
    total_steps += 1
    if result.info.error:
        invalid_count += 1
    print(f"  Round 1 propose | reward={result.reward:.4f} done={result.done}")

    if not result.done:
        # Step 2: accept
        result = env.step(_build_accept_action())
        total_steps += 1
        if result.info.error:
            invalid_count += 1
        print(f"  Round 2 accept  | reward={result.reward:.4f} done={result.done}")

    state = env.state()
    info = result.info

    # Build and persist episode log
    from replicalab.models import EpisodeLog

    episode_log = EpisodeLog(
        episode_id=episode_id,
        seed=state.seed,
        scenario_template=state.scenario_template,
        difficulty=state.difficulty,
        final_state=state,
        transcript=list(state.conversation_history),
        reward_breakdown=info.reward_breakdown,
        total_reward=state.reward,
        rounds_used=state.round_number,
        agreement_reached=info.agreement_reached,
        judge_notes=info.judge_notes or "",
        verdict=info.verdict or "",
        top_failure_reasons=list(info.top_failure_reasons),
        invalid_action_count=invalid_count,
        invalid_action_rate=round(invalid_count / total_steps, 6) if total_steps else 0.0,
    )

    replay_path = write_episode_log(episode_log)
    csv_path, jsonl_path = log_episode_reward(
        episode_id=episode_id,
        seed=state.seed,
        scenario_template=state.scenario_template,
        difficulty=state.difficulty,
        total_reward=state.reward,
        breakdown=info.reward_breakdown,
        rounds_used=state.round_number,
        agreement_reached=info.agreement_reached,
        verdict=info.verdict or "",
        judge_notes=info.judge_notes or "",
    )

    print(f"\n  Verdict: {info.verdict}")
    print(f"  Total reward: {state.reward:.4f}")
    print(f"  Agreement: {info.agreement_reached}")
    print(f"  Invalid actions: {invalid_count}/{total_steps}")
    print(f"\n  Replay JSON:  {replay_path}")
    print(f"  Reward CSV:   {csv_path}")
    print(f"  Reward JSONL: {jsonl_path}")

    env.close()


def main() -> None:
    parser = argparse.ArgumentParser(description="Run a single ReplicaLab episode")
    parser.add_argument("--seed", type=int, default=0)
    parser.add_argument("--scenario", default=DEFAULT_SCENARIO_TEMPLATE)
    parser.add_argument("--difficulty", default=DEFAULT_DIFFICULTY)
    parser.add_argument("--max-rounds", type=int, default=None)
    args = parser.parse_args()
    run_episode(
        seed=args.seed,
        scenario=args.scenario,
        difficulty=args.difficulty,
        max_rounds=args.max_rounds,
    )


if __name__ == "__main__":
    main()