File size: 3,804 Bytes
e918a5c | 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 | #!/usr/bin/env python3
"""
Run the standard pathway agent eval manifest (fixed policy baseline).
Scores via ``env.episode_outcome`` (orchestrator mode). Writes JSON summary.
"""
from __future__ import annotations
import argparse
import json
from pathlib import Path
from typing import Any, Dict, List
from pathway_analysis_env.models import PathwayAction
from pathway_analysis_env.server.analysis import gseapy_available, pydeseq2_available
from pathway_analysis_env.server.pathway_environment import (
DATA_DIR,
PathwayEnvironment,
)
def _run_episode(spec: Dict[str, Any], *, strict: bool) -> Dict[str, Any]:
case_file = spec["case_file"]
if spec.get("requires_pydeseq2") and not pydeseq2_available():
return {
"id": spec["id"],
"skipped": True,
"reason": "pydeseq2_unavailable",
}
if spec.get("requires_gseapy") and not gseapy_available():
return {
"id": spec["id"],
"skipped": True,
"reason": "gseapy_unavailable",
}
from pathway_analysis_env.server.pathway_environment import load_case
case = load_case(case_file)
ref = (case.get("default_contrast") or {}).get("reference")
alt = (case.get("default_contrast") or {}).get("alternate")
env = PathwayEnvironment(case_file=case_file)
env.reset(strict=strict, orchestrator_mode=True)
actions: List[str] = []
def go(kind: str, **kw: Any):
actions.append(kind)
return env.step(PathwayAction(action_type=kind, **kw))
go("understand_experiment_design")
go("inspect_dataset")
o_de = go(
"run_differential_expression",
condition_a=ref,
condition_b=alt,
)
if o_de.metadata and o_de.metadata.get("failure_code"):
return {
"id": spec["id"],
"case_file": case_file,
"passed": False,
"stage": "de",
"failure_code": o_de.metadata.get("failure_code"),
"actions": actions,
}
o_ora = go("run_pathway_enrichment")
if o_ora.metadata and o_ora.metadata.get("failure_code"):
return {
"id": spec["id"],
"case_file": case_file,
"passed": False,
"stage": "ora",
"failure_code": o_ora.metadata.get("failure_code"),
"actions": actions,
}
hyp = spec.get("hypothesis", "")
go("submit_answer", hypothesis=hyp)
outcome = env.episode_outcome or {}
return {
"id": spec["id"],
"case_file": case_file,
"passed": bool(outcome.get("correct")),
"episode_outcome": outcome,
"actions": actions,
"steps": env.state.step_count,
}
def main() -> None:
parser = argparse.ArgumentParser()
parser.add_argument(
"--manifest",
type=Path,
default=DATA_DIR / "eval_manifest.json",
)
parser.add_argument("--strict", action="store_true")
parser.add_argument("--json-out", type=Path, default=None)
args = parser.parse_args()
manifest = json.loads(args.manifest.read_text(encoding="utf-8"))
results = [_run_episode(ep, strict=args.strict) for ep in manifest.get("episodes", [])]
n_pass = sum(1 for r in results if r.get("passed"))
n_run = sum(1 for r in results if not r.get("skipped"))
summary = {
"manifest": str(args.manifest),
"passed": n_pass,
"run": n_run,
"total": len(results),
"pass_rate": (n_pass / n_run) if n_run else 0.0,
"results": results,
}
text = json.dumps(summary, indent=2)
if args.json_out:
args.json_out.parent.mkdir(parents=True, exist_ok=True)
args.json_out.write_text(text + "\n", encoding="utf-8")
print(text)
if __name__ == "__main__":
main()
|