Skip to content

https://arxiv.org/abs/2606.28344. The end of web parsing. The beginning of scalable pixel-native search. link: https://pixelrag.ai/

10.4 KB
Raw Blame
#!/usr/bin/env python3
"""Eval Qwen3-VL-4B on multi-image retrieval-style test data.

Input: ShareGPT JSON produced by prepare_sft_data_multiimage.py
(each example has 6 images + 1 query + 1 golden answer).

Metrics: exact_match / char_accuracy / llm_judge_accuracy (GPT-4.1).
"""

from __future__ import annotations

import argparse
import json
import os
import re
import sys
from concurrent.futures import ThreadPoolExecutor, as_completed
from pathlib import Path
from typing import TypedDict

import torch
from tqdm import tqdm
from transformers import Qwen3VLForConditionalGeneration, AutoProcessor
from qwen_vl_utils import process_vision_info


class EvalResult(TypedDict, total=False):
    query: str
    golden: str
    predicted: str
    image_missing: bool
    missing_paths: list[str]
    n_images: int
    gold_pos: int | None
    gold_in_top6_pos: int | None
    judge_grade: str
    judge_correct: bool


class EMCharMetrics(TypedDict):
    exact_match: float
    char_accuracy: float
    scored: int


class JudgeMetrics(TypedDict):
    llm_judge_accuracy: float
    llm_judge_correct: int
    llm_judge_total: int


_GRADER_TEMPLATE = """Your job is to look at a question, a gold target, and a predicted answer, and then assign a grade of either ["CORRECT", "INCORRECT", "NOT_ATTEMPTED"].
Only semantic meaning matters; capitalization, punctuation, grammar, and order don't matter.
Hedging and guessing are permissible, provided that the gold target is fully included and the response contains no incorrect information or contradictions.
For numerical answers the predicted answer must be correct to the last significant figure in the gold answer. The gold target may contain more information than the question; the predicted answer only needs to contain what the question asks.
Do not punish typos in names if it is clearly the same name.

Here is a new example. Simply reply with either CORRECT, INCORRECT, NOT ATTEMPTED. Don't apologize or correct yourself; we are just grading the answer.
```
Question: {question}
Gold target: {target}
Predicted answer: {predicted_answer}
```

Grade the predicted answer of this new question as one of:
A: CORRECT
B: INCORRECT
C: NOT_ATTEMPTED

Just return the letters "A", "B", or "C", with no text around it."""


def strip_image_tokens(s: str) -> str:
    return s.replace("<image>", "").lstrip("\n ")


def run_inference(
    model: Qwen3VLForConditionalGeneration,
    processor: AutoProcessor,
    examples: list[dict[str, object]],
    device: str,
    desc: str,
    max_new_tokens: int = 128,
    enable_thinking: bool = False,
) -> list[EvalResult]:
    results = []
    for ex in tqdm(examples, desc=desc):
        images = ex.get("images", [])
        # Verify all images exist
        missing = [p for p in images if not os.path.exists(p)]
        if missing:
            results.append(
                {
                    "query": ex.get("_query", ""),
                    "golden": ex.get("_golden", ""),
                    "predicted": "",
                    "image_missing": True,
                    "missing_paths": missing,
                }
            )
            continue

        user_msg = next(m for m in ex["messages"] if m["role"] == "user")
        assistant_msg = next(m for m in ex["messages"] if m["role"] == "assistant")
        query = strip_image_tokens(user_msg["content"])
        golden = assistant_msg["content"].strip()

        content = [{"type": "image", "image": f"file://{p}"} for p in images]
        content.append({"type": "text", "text": query})
        messages = [{"role": "user", "content": content}]

        text = processor.apply_chat_template(
            messages,
            tokenize=False,
            add_generation_prompt=True,
            enable_thinking=enable_thinking,
        )
        image_inputs, video_inputs = process_vision_info(messages)
        inputs = processor(
            text=[text],
            images=image_inputs,
            videos=video_inputs,
            padding=True,
            return_tensors="pt",
        ).to(device)

        with torch.no_grad():
            out_ids = model.generate(
                **inputs, max_new_tokens=max_new_tokens, do_sample=False
            )
        gen_ids = out_ids[0][inputs.input_ids.shape[1] :]
        pred = processor.decode(gen_ids, skip_special_tokens=True).strip()

        results.append(
            {
                "query": query,
                "golden": golden,
                "predicted": pred,
                "n_images": len(images),
                "gold_pos": ex.get("_gold_pos"),
                "gold_in_top6_pos": ex.get("_gold_in_top6_pos"),
            }
        )
    return results


def compute_em_char(results: list[EvalResult]) -> EMCharMetrics:
    correct_em = 0
    char_correct = 0
    char_total = 0
    scored = 0
    for r in results:
        if r.get("image_missing"):
            continue
        scored += 1
        pred = r["predicted"].lower()
        gold = r["golden"].lower()
        if pred == gold:
            correct_em += 1
        if gold or pred:
            matches = sum(1 for a, b in zip(pred, gold) if a == b)
            char_correct += matches
            char_total += max(len(pred), len(gold))
    return {
        "exact_match": correct_em / scored if scored else 0.0,
        "char_accuracy": char_correct / char_total if char_total else 0.0,
        "scored": scored,
    }


def grade_with_gpt(
    results: list[EvalResult], model: str, concurrency: int = 16
) -> JudgeMetrics:
    from openai import OpenAI

    client = OpenAI()

    def _grade(idx, r):
        try:
            resp = client.chat.completions.create(
                model=model,
                messages=[
                    {
                        "role": "user",
                        "content": _GRADER_TEMPLATE.format(
                            question=r["query"],
                            target=r["golden"],
                            predicted_answer=r["predicted"] or "(no answer)",
                        ),
                    }
                ],
                max_tokens=5,
                temperature=0,
            )
            grade = resp.choices[0].message.content.strip()
            is_correct = bool(re.match(r"^\s*A\b", grade))
            return idx, is_correct, grade
        except Exception as e:
            return idx, False, f"ERR:{e}"

    scored = [(i, r) for i, r in enumerate(results) if not r.get("image_missing")]
    n = len(scored)
    verdicts = [("", False)] * len(results)
    correct = 0
    with ThreadPoolExecutor(max_workers=concurrency) as pool:
        futures = [pool.submit(_grade, i, r) for i, r in scored]
        for fut in tqdm(as_completed(futures), total=n, desc="GPT judge"):
            idx, is_correct, grade = fut.result()
            verdicts[idx] = (grade, is_correct)
            if is_correct:
                correct += 1

    for i, (grade, is_correct) in enumerate(verdicts):
        results[i]["judge_grade"] = grade
        results[i]["judge_correct"] = is_correct
    return {
        "llm_judge_accuracy": correct / n if n else 0.0,
        "llm_judge_correct": correct,
        "llm_judge_total": n,
    }


def main() -> None:
    p = argparse.ArgumentParser()
    p.add_argument("--model", default="Qwen/Qwen3-VL-4B-Instruct")
    p.add_argument("--adapter", default=None)
    p.add_argument(
        "--test-json",
        required=True,
        help="Path to <dataset_dir>/test.json from prepare_sft_data_multiimage.py",
    )
    p.add_argument("--n-examples", type=int, default=500)
    p.add_argument("--max-new-tokens", type=int, default=128)
    p.add_argument("--thinking", action="store_true")
    p.add_argument("--device", default="cuda:0")
    p.add_argument("--judge", action="store_true", default=True)
    p.add_argument("--no-judge", dest="judge", action="store_false")
    p.add_argument("--judge-model", default="gpt-4.1-2025-04-14")
    p.add_argument("--judge-concurrency", type=int, default=16)
    p.add_argument("--tag", required=True)
    p.add_argument(
        "--output-dir",
        default="/scratch/users/zwcolin/cxr_embeds/cxr_embedding/sft/eval_out",
    )
    args = p.parse_args()

    if args.judge and not os.environ.get("OPENAI_API_KEY"):
        print("ERROR: --judge set but OPENAI_API_KEY not in env.", file=sys.stderr)
        sys.exit(1)

    print(f"Loading base model: {args.model}")
    model = Qwen3VLForConditionalGeneration.from_pretrained(
        args.model,
        torch_dtype=torch.bfloat16,
        device_map=args.device,
    )
    if args.adapter:
        from peft import PeftModel

        print(f"Loading LoRA adapter: {args.adapter}")
        model = PeftModel.from_pretrained(model, args.adapter)
        model = model.merge_and_unload()
    model.eval()

    processor = AutoProcessor.from_pretrained(args.model)

    with open(args.test_json) as f:
        examples = json.load(f)
    if args.n_examples > 0:
        examples = examples[: args.n_examples]
    print(f"Loaded {len(examples)} multi-image examples")

    if args.thinking and args.max_new_tokens < 512:
        args.max_new_tokens = 512

    results = run_inference(
        model,
        processor,
        examples,
        args.device,
        desc=f"eval[{args.tag}]",
        max_new_tokens=args.max_new_tokens,
        enable_thinking=args.thinking,
    )

    metrics = compute_em_char(results)
    print(f"\n=== {args.tag} ===")
    print(f"  scored:        {metrics['scored']} / {len(results)}")
    print(f"  exact_match:   {metrics['exact_match']:.4f}")
    print(f"  char_accuracy: {metrics['char_accuracy']:.4f}")

    if args.judge:
        judge_metrics = grade_with_gpt(
            results, args.judge_model, args.judge_concurrency
        )
        metrics.update(judge_metrics)
        print(
            f"  llm_judge:     {metrics['llm_judge_accuracy']:.4f} "
            f"({metrics['llm_judge_correct']}/{metrics['llm_judge_total']})"
        )

    Path(args.output_dir).mkdir(parents=True, exist_ok=True)
    out = {
        "tag": args.tag,
        "model": args.model,
        "adapter": args.adapter,
        "test_json": args.test_json,
        "n_examples": len(results),
        "judge_model": args.judge_model if args.judge else None,
        "metrics": metrics,
        "results": results,
    }
    fname = f"eval_{args.tag}_multiimage_n{len(results)}.json"
    fpath = os.path.join(args.output_dir, fname)
    with open(fpath, "w") as f:
        json.dump(out, f, ensure_ascii=False, indent=2)
    print(f"\nSaved: {fpath}")


if __name__ == "__main__":
    main()