code/rl_inference/test_classifier_on_gpt5_outputs.py

import argparse
import glob
import json
import os
import traceback
import urllib.error
import urllib.request
from collections import defaultdict
from datetime import datetime
from typing import Any, DefaultDict, Dict, List

import dspy
from tqdm import tqdm


DEFAULT_API_BASE = "http://172.16.34.22:8040/v1"
DEFAULT_MODEL_PATH = (
    "/home/mshahidul/readctrl/code/text_classifier/"
    "dspy_model/vllm-Meta-Llama-3.1-8B-Instruct_teacher-gpt5_v1/model.json"
)
DEFAULT_OUTPUT_DIR = "/home/mshahidul/readctrl/code/rl_inference/test_result"

VALID_LABELS = {
    "low_health_literacy",
    "intermediate_health_literacy",
    "proficient_health_literacy",
}


class HealthLiteracySignature(dspy.Signature):
    generated_text = dspy.InputField(
        desc="A version of the source text rewritten for a specific audience."
    )
    literacy_label = dspy.OutputField(
        desc=(
            "Classification: low_health_literacy (simple words, no jargon), "
            "intermediate_health_literacy (moderate technicality), or "
            "proficient_health_literacy (highly technical/original level)."
        )
    )


class HealthLiteracyClassifier(dspy.Module):
    def __init__(self):
        super().__init__()
        self.classifier = dspy.ChainOfThought(HealthLiteracySignature)

    def forward(self, generated_text):
        return self.classifier(generated_text=generated_text)


def parse_args() -> argparse.Namespace:
    parser = argparse.ArgumentParser(
        description="Evaluate GPT output files with saved DSPy health literacy classifier."
    )
    parser.add_argument("--model-path", default=DEFAULT_MODEL_PATH)
    parser.add_argument(
        "--input-path",
        default="",
        help=(
            "Path to GPT output JSONL (e.g. gpt5_inference_all_*.jsonl). "
            "If omitted, auto-select latest file in test_result."
        ),
    )
    parser.add_argument(
        "--api-base",
        default=os.environ.get("VLLM_API_BASE", DEFAULT_API_BASE),
    )
    parser.add_argument("--output-dir", default=DEFAULT_OUTPUT_DIR)
    parser.add_argument(
        "--max-samples",
        type=int,
        default=-1,
        help="Use -1 for all valid rows.",
    )
    parser.add_argument(
        "--provide-traceback",
        action="store_true",
        help="Print full traceback if runtime error happens.",
    )
    return parser.parse_args()


def resolve_input_path(input_path: str, search_dir: str) -> str:
    if input_path and os.path.exists(input_path):
        return input_path
    if input_path:
        raise FileNotFoundError(f"Input file not found: {input_path}")

    candidates = sorted(glob.glob(os.path.join(search_dir, "gpt5_inference_all_*.jsonl")), key=os.path.getmtime)
    if not candidates:
        # Fallback: allow evaluating model-specific inference outputs too.
        candidates = sorted(
            glob.glob(os.path.join(search_dir, "gpt5_inference_*_*.jsonl")),
            key=os.path.getmtime,
        )
    if not candidates:
        raise FileNotFoundError(
            "No GPT output file found. Expected pattern: "
            f"{search_dir}/gpt5_inference_all_*.jsonl "
            "or gpt5_inference_*_*.jsonl"
        )
    return candidates[-1]


def check_api_base(api_base: str) -> None:
    models_url = api_base.rstrip("/") + "/models"
    req = urllib.request.Request(models_url, method="GET")
    try:
        with urllib.request.urlopen(req, timeout=5) as resp:
            if resp.status >= 400:
                raise RuntimeError(
                    f"Endpoint reachable but unhealthy: {models_url} (status={resp.status})"
                )
    except urllib.error.URLError as exc:
        raise ConnectionError(
            "Cannot reach OpenAI-compatible endpoint. "
            f"api_base={api_base}. "
            "Start your vLLM server or pass correct --api-base."
        ) from exc


def load_compiled_classifier(path: str):
    if hasattr(dspy, "load"):
        try:
            return dspy.load(path)
        except Exception:
            pass

    classifier = HealthLiteracyClassifier()
    try:
        classifier.load(path)
    except Exception as exc:
        raise RuntimeError(f"Failed to load compiled model from {path}") from exc
    return classifier


def normalize_pred_label(pred_obj: Any) -> str:
    if not pred_obj or not hasattr(pred_obj, "literacy_label"):
        return ""
    return str(pred_obj.literacy_label).strip().lower()


def load_eval_items(path: str) -> List[Dict[str, Any]]:
    items: List[Dict[str, Any]] = []
    with open(path, "r", encoding="utf-8") as f:
        for line_no, line in enumerate(f, start=1):
            if not line.strip():
                continue
            row = json.loads(line)
            gold_label = str(row.get("gold_label", "")).strip()
            generated_text = str(row.get("generated_text", "")).strip()
            err_msg = str(row.get("error", "")).strip()

            if gold_label not in VALID_LABELS:
                continue
            if err_msg:
                continue
            if not generated_text:
                continue

            items.append(
                {
                    "line_no": line_no,
                    "model": str(row.get("model", "")).strip() or "unknown_model",
                    "row_index": row.get("row_index"),
                    "doc_id": row.get("doc_id"),
                    "gold_label": gold_label,
                    "generated_text": generated_text,
                }
            )
    return items


def main() -> None:
    args = parse_args()
    args.input_path = resolve_input_path(args.input_path, args.output_dir)

    if not os.path.exists(args.model_path):
        raise FileNotFoundError(f"Model file not found: {args.model_path}")

    try:
        check_api_base(args.api_base)
        lm = dspy.LM(
            model="openai/dspy",
            api_base=args.api_base,
            api_key="EMPTY",
            temperature=0.0,
        )
        dspy.configure(lm=lm)
        classifier = load_compiled_classifier(args.model_path)
        print(f"[INFO] Using input file: {args.input_path}")

        eval_items = load_eval_items(args.input_path)
        if args.max_samples > 0:
            eval_items = eval_items[: args.max_samples]
        if not eval_items:
            raise RuntimeError("No valid rows found for evaluation.")

        results: List[Dict[str, Any]] = []
        model_total: DefaultDict[str, int] = defaultdict(int)
        model_correct: DefaultDict[str, int] = defaultdict(int)

        for item in tqdm(eval_items, desc="Classifying"):
            pred = classifier(generated_text=item["generated_text"])
            pred_label = normalize_pred_label(pred)
            is_correct = item["gold_label"] in pred_label

            model_name = item["model"]
            model_total[model_name] += 1
            model_correct[model_name] += int(is_correct)

            results.append(
                {
                    "line_no": item["line_no"],
                    "model": model_name,
                    "row_index": item["row_index"],
                    "doc_id": item["doc_id"],
                    "gold_label": item["gold_label"],
                    "pred_label": pred_label,
                    "is_correct": is_correct,
                }
            )

        total = len(results)
        correct = sum(1 for r in results if r["is_correct"])
        overall_accuracy = correct / total if total else 0.0

        per_model: Dict[str, Dict[str, Any]] = {}
        for model_name in sorted(model_total.keys()):
            m_total = model_total[model_name]
            m_correct = model_correct[model_name]
            per_model[model_name] = {
                "total_samples": m_total,
                "correct_samples": m_correct,
                "accuracy_score": (m_correct / m_total) if m_total else 0.0,
            }

        ts = datetime.now().strftime("%Y%m%d_%H%M%S")
        os.makedirs(args.output_dir, exist_ok=True)
        summary_path = os.path.join(args.output_dir, f"classifier_eval_gpt5_{ts}.json")
        details_path = os.path.join(args.output_dir, f"classifier_eval_gpt5_{ts}.jsonl")

        summary_obj = {
            "model_path": args.model_path,
            "input_path": args.input_path,
            "api_base": args.api_base,
            "total_samples": total,
            "correct_samples": correct,
            "accuracy_score": overall_accuracy,
            "per_model": per_model,
            "details_path": details_path,
        }

        with open(summary_path, "w", encoding="utf-8") as f:
            json.dump(summary_obj, f, indent=2, ensure_ascii=False)

        with open(details_path, "w", encoding="utf-8") as f:
            for record in results:
                f.write(json.dumps(record, ensure_ascii=False) + "\n")

        print(json.dumps(summary_obj, indent=2, ensure_ascii=False))
        print(f"[DONE] Summary saved: {summary_path}")
        print(f"[DONE] Details saved: {details_path}")

    except Exception as exc:
        print(f"[error] {type(exc).__name__}: {exc}")
        if args.provide_traceback:
            traceback.print_exc()
        raise


if __name__ == "__main__":
    main()