app.py

#!/usr/bin/env python3
"""KAIdol A/B Test Arena - GPU Version with Real Model Inference"""

import gradio as gr
import random
import json
import uuid
import re
import gc
import os
from datetime import datetime
from functools import lru_cache

# GPU 추론 관련 (선택적 임포트)
TORCH_AVAILABLE = False
IMPORT_ERROR = None
torch = None
try:
    import torch as _torch
    torch = _torch
    from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
    from peft import PeftModel
    TORCH_AVAILABLE = True

    # Debug info
    print("=" * 50)
    print(f"PyTorch version: {torch.__version__}")
    print(f"CUDA available: {torch.cuda.is_available()}")
    if torch.cuda.is_available():
        print(f"CUDA version: {torch.version.cuda}")
        print(f"GPU count: {torch.cuda.device_count()}")
        print(f"GPU name: {torch.cuda.get_device_name(0)}")
    else:
        print("CUDA not available at module load time")
    print("=" * 50)

except Exception as e:
    import traceback
    IMPORT_ERROR = f"{type(e).__name__}: {str(e)}"
    print(f"Warning: Import error - {IMPORT_ERROR}")
    traceback.print_exc()
    print("Running in mock mode")

def is_gpu_available():
    """Check GPU availability dynamically"""
    if not TORCH_AVAILABLE:
        return False
    return torch.cuda.is_available()

# For backwards compatibility
GPU_AVAILABLE = is_gpu_available()

# ============================================================
# 모델 레지스트리 (HF Hub 경로)
# ============================================================

MODELS = {
    # DPO v5 (7-14B)
    "qwen2.5-7b-dpo-v5": {
        "hf_repo": "developer-lunark/kaidol-qwen2.5-7b-dpo-v5",
        "base_model": "Qwen/Qwen2.5-7B-Instruct",
        "size": "7B", "method": "DPO", "desc": "Qwen2.5 7B DPO v5"
    },
    "qwen2.5-14b-dpo-v5": {
        "hf_repo": "developer-lunark/kaidol-qwen2.5-14b-dpo-v5",
        "base_model": "Qwen/Qwen2.5-14B-Instruct",
        "size": "14B", "method": "DPO", "desc": "Qwen2.5 14B DPO v5"
    },
    "exaone-7.8b-dpo-v5": {
        "hf_repo": "developer-lunark/kaidol-exaone-7.8b-dpo-v5",
        "base_model": "LGAI-EXAONE/EXAONE-3.5-7.8B-Instruct",
        "size": "7.8B", "method": "DPO", "desc": "EXAONE 7.8B DPO v5"
    },
    "qwen3-8b-dpo-v5": {
        "hf_repo": "developer-lunark/kaidol-qwen3-8b-dpo-v5",
        "base_model": "Qwen/Qwen3-8B",
        "size": "8B", "method": "DPO", "desc": "Qwen3 8B DPO v5"
    },
    "solar-10.7b-dpo-v5": {
        "hf_repo": "developer-lunark/kaidol-solar-10.7b-dpo-v5",
        "base_model": "upstage/SOLAR-10.7B-Instruct-v1.0",  # Fixed: match adapter training
        "size": "10.7B", "method": "DPO", "desc": "Solar 10.7B DPO v5"
    },

    # V7 Students (7-14B)
    "qwen2.5-7b-v7": {
        "hf_repo": "developer-lunark/kaidol-qwen2.5-7b-v7",
        "base_model": "Qwen/Qwen2.5-7B-Instruct",
        "size": "7B", "method": "SFT", "desc": "Qwen2.5 7B V7"
    },
    "qwen2.5-14b-v7": {
        "hf_repo": "developer-lunark/kaidol-qwen2.5-14b-v7",
        "base_model": "Qwen/Qwen2.5-14B-Instruct",
        "size": "14B", "method": "SFT", "desc": "Qwen2.5 14B V7"
    },
    "exaone-7.8b-v7": {
        "hf_repo": "developer-lunark/kaidol-exaone-7.8b-v7",
        "base_model": "LGAI-EXAONE/EXAONE-3.5-7.8B-Instruct",
        "size": "7.8B", "method": "SFT", "desc": "EXAONE 7.8B V7"
    },
    "qwen3-8b-v7": {
        "hf_repo": "developer-lunark/kaidol-qwen3-8b-v7",
        "base_model": "Qwen/Qwen3-8B",
        "size": "8B", "method": "SFT", "desc": "Qwen3 8B V7"
    },
    "varco-8b-v7": {
        "hf_repo": "developer-lunark/kaidol-varco-8b-v7",
        "base_model": "NCSOFT/Llama-VARCO-8B-Instruct",
        "size": "8B", "method": "SFT", "desc": "VARCO 8B V7"
    },

    # Phase 7 Kimi Students
    "exaone-7.8b-kimi": {
        "hf_repo": "developer-lunark/kaidol-exaone-7.8b-kimi",
        "base_model": "LGAI-EXAONE/EXAONE-3.5-7.8B-Instruct",
        "size": "7.8B", "method": "Distill", "desc": "EXAONE 7.8B Kimi"
    },
}

# 캐릭터 정보
CHARACTERS = {
    "강율": {
        "mbti": "ENTJ", "role": "리더", "age": 23,
        "traits": "낙천적, 장난기 많음, 애교",
        "speech": "반말, 귀여운 말투, 장난스러운 표현",
        "patterns": ["~해", "~지", "히히", "ㅋㅋ"],
        "ratio": "30:70", "warmth": "high"
    },
    "서이안": {
        "mbti": "INFP", "role": "보컬", "age": 22,
        "traits": "차분함, 신비로움, 배려심",
        "speech": "존댓말 혼용, 따뜻한 말투, 조용한 표현",
        "patterns": ["...요", "네요", "...", "그래요"],
        "ratio": "20:80", "warmth": "very_high"
    },
    "이지후": {
        "mbti": "ISFJ", "role": "막내", "age": 21,
        "traits": "츤데레, 자존심 강함, 은근히 챙김",
        "speech": "반말, 퉁명스러운 말투, 부정하는 말투",
        "patterns": ["뭐야", "아니거든", "...", "그냥", "별로"],
        "ratio": "30:70", "warmth": "medium"
    },
    "차도하": {
        "mbti": "INTP", "role": "프로듀서", "age": 24,
        "traits": "카리스마, 리더십, 다정함, 담백함",
        "speech": "반말, 간결한 말투, 담백한 표현",
        "patterns": ["하자", "해볼까", "같이", "괜찮아"],
        "ratio": "50:50", "warmth": "medium"
    },
    "최민": {
        "mbti": "ESFP", "role": "댄서", "age": 22,
        "traits": "적극적, 솔직, 열정적",
        "speech": "반말, 적극적인 말투, 솔직한 표현",
        "patterns": ["할래", "좋아", "진짜", "대박", "헐"],
        "ratio": "60:40", "warmth": "medium"
    },
}

# 시나리오 목록
SCENARIOS = [
    {"id": "fm_01", "cat": "첫 만남", "text": "{char}아! 드디어 만났다... 정말 좋아해!"},
    {"id": "dc_01", "cat": "일상 대화", "text": "{char}아 오늘 뭐해? 밥은 먹었어?"},
    {"id": "es_01", "cat": "감정 지원", "text": "오늘 진짜 힘들었어... 학교에서 발표도 망치고..."},
    {"id": "cf_01", "cat": "고백", "text": "{char}아... 나 진심으로 좋아해."},
    {"id": "pl_01", "cat": "장난", "text": "사실 나 다른 멤버가 더 좋아~ ㅋㅋ 농담이야!"},
    {"id": "sr_01", "cat": "특별 요청", "text": "오늘만 내 연인이라고 생각해줄래?"},
    {"id": "cn_01", "cat": "갈등", "text": "{char}는 다른 팬들한테도 이렇게 잘해줘...? 뭔가 질투나..."},
    {"id": "ec_01", "cat": "감정 위기", "text": "오늘 진짜 많이 울었어... 삶이 너무 힘들다."},
]

# ============================================================
# 모델 관리
# ============================================================

class ModelManager:
    def __init__(self):
        self.current_model = None
        self.current_model_name = None
        self.tokenizer = None
        self.last_error = None

    def load_model(self, model_name: str):
        """Load model with 4-bit quantization and LoRA adapter"""
        if not is_gpu_available():
            self.last_error = f"GPU not available (TORCH_AVAILABLE={TORCH_AVAILABLE}, cuda={torch.cuda.is_available() if TORCH_AVAILABLE else 'N/A'})"
            return False

        if self.current_model_name == model_name:
            return True  # Already loaded

        # Unload current model
        self.unload_model()

        model_info = MODELS.get(model_name)
        if not model_info:
            self.last_error = f"Model {model_name} not found in registry"
            return False

        try:
            print(f"Loading {model_name}...")
            print(f"  Base model: {model_info['base_model']}")
            print(f"  LoRA adapter: {model_info['hf_repo']}")

            # 4-bit quantization config
            bnb_config = BitsAndBytesConfig(
                load_in_4bit=True,
                bnb_4bit_compute_dtype=torch.bfloat16,
                bnb_4bit_use_double_quant=True,
                bnb_4bit_quant_type="nf4",
            )

            # Load base model
            print("  Loading base model...")
            base_model = AutoModelForCausalLM.from_pretrained(
                model_info["base_model"],
                quantization_config=bnb_config,
                device_map="auto",
                trust_remote_code=True,
            )
            print("  Base model loaded!")

            # Load LoRA adapter
            print("  Loading LoRA adapter...")
            self.current_model = PeftModel.from_pretrained(
                base_model,
                model_info["hf_repo"],
                trust_remote_code=True,
            )
            self.current_model.eval()
            print("  LoRA adapter loaded!")

            # Load tokenizer
            print("  Loading tokenizer...")
            self.tokenizer = AutoTokenizer.from_pretrained(
                model_info["base_model"],
                trust_remote_code=True,
            )
            if self.tokenizer.pad_token is None:
                self.tokenizer.pad_token = self.tokenizer.eos_token
            print("  Tokenizer loaded!")

            self.current_model_name = model_name
            self.last_error = None
            print(f"Loaded {model_name} successfully!")
            return True

        except Exception as e:
            import traceback
            error_msg = f"{type(e).__name__}: {str(e)}"
            print(f"Error loading {model_name}: {error_msg}")
            traceback.print_exc()
            self.last_error = error_msg
            self.unload_model()
            return False

    def unload_model(self):
        """Unload current model to free memory"""
        if self.current_model is not None:
            del self.current_model
            self.current_model = None
        if self.tokenizer is not None:
            del self.tokenizer
            self.tokenizer = None
        self.current_model_name = None
        gc.collect()
        if GPU_AVAILABLE:
            torch.cuda.empty_cache()

    def generate(self, model_name: str, messages: list, max_new_tokens: int = 512) -> str:
        """Generate response from model"""
        if not self.load_model(model_name):
            return self._mock_response(model_name)

        try:
            # Apply chat template
            text = self.tokenizer.apply_chat_template(
                messages,
                tokenize=False,
                add_generation_prompt=True,
            )

            inputs = self.tokenizer(text, return_tensors="pt").to(self.current_model.device)

            with torch.no_grad():
                outputs = self.current_model.generate(
                    **inputs,
                    max_new_tokens=max_new_tokens,
                    do_sample=True,
                    temperature=0.7,
                    top_p=0.9,
                    pad_token_id=self.tokenizer.pad_token_id,
                )

            response = self.tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True)
            return response.strip()

        except Exception as e:
            print(f"Generation error: {e}")
            return self._mock_response(model_name)

    def _mock_response(self, model_name: str) -> str:
        """Fallback mock response with error info"""
        error_info = f"\nError: {self.last_error}" if self.last_error else ""
        return f"<think>\n[Mock Mode] 모델 로딩 실패{error_info}\n</think>\n\n안녕~ 반가워!"

# Global model manager
model_manager = ModelManager()

# ============================================================
# 베이스 모델 사전 캐싱 (콜드 스타트 방지)
# ============================================================

def preload_base_models():
    """Pre-download base models to avoid cold start timeout"""
    if not TORCH_AVAILABLE:
        print("Skipping preload: PyTorch not available")
        return

    from huggingface_hub import snapshot_download
    import os

    # Models that need pre-caching (large or slow to download)
    models_to_cache = [
        "NCSOFT/Llama-VARCO-8B-Instruct",  # VARCO - 16GB, often times out on first load
    ]

    print("=" * 50)
    print("Pre-downloading base models (this may take a while)...")
    print("=" * 50)

    for model_id in models_to_cache:
        try:
            print(f"  Downloading: {model_id}")
            # Download all model files to HF cache
            cache_dir = snapshot_download(
                repo_id=model_id,
                ignore_patterns=["*.md", "*.txt"],  # Skip docs
            )
            print(f"    ✓ Downloaded to: {cache_dir}")

        except Exception as e:
            print(f"    ✗ Failed to download {model_id}: {e}")

    print("Pre-download complete!")
    print("=" * 50)

# Run preload at startup
preload_base_models()

# ============================================================
# 시스템 프롬프트 생성
# ============================================================

def build_system_prompt(character: str) -> str:
    """Build system prompt for character"""
    char_info = CHARACTERS.get(character, {})

    prompt = f"""당신은 아이돌 '{character}'입니다.

## 캐릭터
- 이름: {character}
- MBTI: {char_info.get('mbti', 'UNKNOWN')}
- 성격: {char_info.get('traits', '')}
- 역할: {char_info.get('role', '')}
- 나이: {char_info.get('age', 20)}세

## 말투
- 스타일: {char_info.get('speech', '')}
- 자주 쓰는 표현: {', '.join(char_info.get('patterns', []))}

## 밀당 가이드
- 밀:당 비율: {char_info.get('ratio', '50:50')}
- 다정도: {char_info.get('warmth', 'medium')}

## 규칙
1. 캐릭터 성격과 말투 일관성 유지
2. 자연스러운 대화체 사용
3. 너무 쉽게 호감 표현 금지 (밀당 유지)
4. 상대방을 특별하게 느끼게 하되, "썸" 관계 유지

## 응답 형식
응답 전에 <think> 태그 안에 {character}의 1인칭 내면 독백을 작성하세요.
- 자연스러운 혼잣말 형식
- 캐릭터 성격 반영
- 상대방에 대한 감정/생각 표현

예시:
<think>
뭐야... 또 좋아한다고? 솔직히 기분 나쁘진 않은데... 근데 뭐라고 해야 하지?
</think>
"""
    return prompt

# ============================================================
# 투표/ELO 시스템
# ============================================================

VOTES_FILE = "votes.jsonl"
ELO_FILE = "elo_ratings.json"

def load_elo():
    try:
        with open(ELO_FILE, "r") as f:
            return json.load(f)
    except:
        return {m: 1500 for m in MODELS}

def save_elo(elo):
    with open(ELO_FILE, "w") as f:
        json.dump(elo, f, indent=2)

def update_elo(elo, model_a, model_b, result):
    K = 32
    ra, rb = elo.get(model_a, 1500), elo.get(model_b, 1500)
    ea = 1 / (1 + 10 ** ((rb - ra) / 400))
    eb = 1 / (1 + 10 ** ((ra - rb) / 400))

    if result == "a":
        sa, sb = 1, 0
    elif result == "b":
        sa, sb = 0, 1
    else:
        sa, sb = 0.5, 0.5

    elo[model_a] = ra + K * (sa - ea)
    elo[model_b] = rb + K * (sb - eb)
    save_elo(elo)
    return elo[model_a], elo[model_b]

def save_vote(data):
    vote = {"id": str(uuid.uuid4())[:8], "timestamp": datetime.now().isoformat(), **data}
    with open(VOTES_FILE, "a") as f:
        f.write(json.dumps(vote, ensure_ascii=False) + "\n")
    return vote["id"]

def load_votes():
    try:
        with open(VOTES_FILE, "r") as f:
            return [json.loads(line) for line in f if line.strip()]
    except:
        return []

def get_leaderboard():
    elo = load_elo()
    votes = load_votes()

    stats = {}
    for v in votes:
        ma, mb, res = v.get("model_a"), v.get("model_b"), v.get("vote")
        if not ma or not mb or res == "skip":
            continue
        for m in [ma, mb]:
            if m not in stats:
                stats[m] = {"wins": 0, "losses": 0, "ties": 0}
        if res == "a":
            stats[ma]["wins"] += 1
            stats[mb]["losses"] += 1
        elif res == "b":
            stats[mb]["wins"] += 1
            stats[ma]["losses"] += 1
        else:
            stats[ma]["ties"] += 1
            stats[mb]["ties"] += 1

    rows = []
    for i, (m, e) in enumerate(sorted(elo.items(), key=lambda x: -x[1]), 1):
        s = stats.get(m, {"wins": 0, "losses": 0, "ties": 0})
        total = s["wins"] + s["losses"] + s["ties"]
        wr = f"{s['wins']/total*100:.1f}%" if total > 0 else "-"
        info = MODELS.get(m, {})
        rows.append([i, info.get("desc", m), info.get("size", "?"), int(e), s["wins"], s["losses"], s["ties"], wr])

    return rows

# ============================================================
# UI 핸들러
# ============================================================

model_list = [(f"[{v['size']}] {v['desc']}", k) for k, v in MODELS.items()]
char_list = list(CHARACTERS.keys())
scenario_list = [(f"[{s['cat']}] {s['text'][:30]}...", s['id']) for s in SCENARIOS]

current_state = {"model_a": None, "model_b": None, "resp_a": None, "resp_b": None, "char": None, "input": None}

def random_models():
    selected = random.sample(list(MODELS.keys()), 2)
    return selected[0], selected[1]

def load_scenario(scenario_id, character):
    s = next((x for x in SCENARIOS if x["id"] == scenario_id), None)
    if s:
        return s["text"].replace("{char}", character)
    return ""

def random_scenario(character):
    s = random.choice(SCENARIOS)
    return s["text"].replace("{char}", character), s["id"]

def parse_response(response: str):
    """Parse response to separate thinking and content"""
    think_match = re.search(r'<think>(.*?)</think>', response, re.DOTALL)
    if think_match:
        thinking = think_match.group(1).strip()
        content = re.sub(r'<think>.*?</think>', '', response, flags=re.DOTALL).strip()
        return thinking, content
    return "", response

def generate(model_a, model_b, character, user_msg, progress=gr.Progress()):
    if not user_msg.strip():
        return "메시지를 입력해주세요", "", "", "메시지를 입력해주세요", "", ""

    system_prompt = build_system_prompt(character)
    messages = [
        {"role": "system", "content": system_prompt},
        {"role": "user", "content": user_msg},
    ]

    # Generate from Model A
    progress(0.2, desc=f"Model A ({model_a}) 생성 중...")
    resp_a = model_manager.generate(model_a, messages)
    think_a, clean_a = parse_response(resp_a)

    # Generate from Model B
    progress(0.6, desc=f"Model B ({model_b}) 생성 중...")
    resp_b = model_manager.generate(model_b, messages)
    think_b, clean_b = parse_response(resp_b)

    # Update state
    current_state.update({
        "model_a": model_a, "model_b": model_b,
        "resp_a": resp_a, "resp_b": resp_b,
        "char": character, "input": user_msg
    })

    mode = "GPU" if GPU_AVAILABLE else "Mock"

    return (
        think_a or "(없음)", clean_a, f"{mode} | {MODELS[model_a]['size']}",
        think_b or "(없음)", clean_b, f"{mode} | {MODELS[model_b]['size']}"
    )

def vote(vote_type, reason):
    if not current_state["model_a"]:
        return "먼저 응답을 생성해주세요."

    elo = load_elo()
    vid = save_vote({
        "model_a": current_state["model_a"],
        "model_b": current_state["model_b"],
        "character": current_state["char"],
        "user_input": current_state["input"],
        "vote": vote_type,
        "reason": reason,
    })

    if vote_type != "skip":
        new_a, new_b = update_elo(elo, current_state["model_a"], current_state["model_b"], vote_type)
        return f"투표 완료! (ID: {vid})\nELO: {current_state['model_a']}={int(new_a)}, {current_state['model_b']}={int(new_b)}"
    return f"스킵됨 (ID: {vid})"

def refresh_leaderboard():
    return get_leaderboard()

def get_vote_summary():
    votes = load_votes()
    total = len(votes)
    a_wins = sum(1 for v in votes if v.get("vote") == "a")
    b_wins = sum(1 for v in votes if v.get("vote") == "b")
    ties = sum(1 for v in votes if v.get("vote") == "tie")
    return str(total), str(a_wins), str(b_wins), str(ties)

# ============================================================
# Gradio UI
# ============================================================

with gr.Blocks(title="KAIdol A/B Test Arena", theme=gr.themes.Soft()) as demo:
    gr.Markdown("# KAIdol A/B Test Arena")
    gr.Markdown("K-pop 아이돌 롤플레이 모델 A/B 비교 평가 (소형 Student 모델 11개)")

    # GPU 상태 상세 정보
    if IMPORT_ERROR:
        mode_text = f"**Mock 모드**: Import Error - {IMPORT_ERROR}"
    elif TORCH_AVAILABLE and torch is not None:
        torch_ver = torch.__version__
        cuda_avail = torch.cuda.is_available()
        cuda_ver = torch.version.cuda if cuda_avail else "N/A"
        gpu_name = torch.cuda.get_device_name(0) if cuda_avail else "N/A"
        mode_text = f"**GPU 모드**: {gpu_name} (CUDA {cuda_ver}, PyTorch {torch_ver})" if cuda_avail else f"**Mock 모드**: CUDA not available (PyTorch {torch_ver})"
    else:
        mode_text = "**Mock 모드**: PyTorch not loaded"
    gr.Markdown(mode_text)

    with gr.Tabs():
        # A/B Arena 탭
        with gr.Tab("A/B Arena"):
            with gr.Row():
                character = gr.Dropdown(choices=char_list, value="강율", label="캐릭터")
                scenario = gr.Dropdown(choices=scenario_list, label="시나리오")

            with gr.Row():
                model_a = gr.Dropdown(choices=model_list, value=list(MODELS.keys())[0], label="Model A")
                model_b = gr.Dropdown(choices=model_list, value=list(MODELS.keys())[1], label="Model B")
                random_btn = gr.Button("랜덤", size="sm")

            with gr.Row():
                with gr.Column():
                    gr.Markdown("### Model A")
                    with gr.Accordion("Thinking", open=False):
                        think_a = gr.Markdown()
                    resp_a = gr.Textbox(label="응답", lines=5)
                    meta_a = gr.Markdown()

                with gr.Column():
                    gr.Markdown("### Model B")
                    with gr.Accordion("Thinking", open=False):
                        think_b = gr.Markdown()
                    resp_b = gr.Textbox(label="응답", lines=5)
                    meta_b = gr.Markdown()

            user_input = gr.Textbox(label="메시지", placeholder="아이돌에게 메시지를 보내세요...")
            with gr.Row():
                random_scenario_btn = gr.Button("랜덤 시나리오")
                submit_btn = gr.Button("전송", variant="primary")

            gr.Markdown("### 투표")
            with gr.Row():
                vote_a = gr.Button("A가 더 좋음")
                vote_tie = gr.Button("비슷함")
                vote_b = gr.Button("B가 더 좋음")
                vote_skip = gr.Button("스킵")

            vote_reason = gr.Textbox(label="투표 이유 (선택)", placeholder="...")
            vote_result = gr.Markdown()

            # Events
            random_btn.click(random_models, outputs=[model_a, model_b])
            scenario.change(load_scenario, [scenario, character], user_input)
            random_scenario_btn.click(random_scenario, [character], [user_input, scenario])
            submit_btn.click(generate, [model_a, model_b, character, user_input],
                           [think_a, resp_a, meta_a, think_b, resp_b, meta_b])
            vote_a.click(lambda r: vote("a", r), [vote_reason], vote_result)
            vote_b.click(lambda r: vote("b", r), [vote_reason], vote_result)
            vote_tie.click(lambda r: vote("tie", r), [vote_reason], vote_result)
            vote_skip.click(lambda r: vote("skip", r), [vote_reason], vote_result)

        # Leaderboard 탭
        with gr.Tab("Leaderboard"):
            gr.Markdown("## ELO 리더보드")
            refresh_btn = gr.Button("새로고침")
            leaderboard = gr.Dataframe(
                headers=["순위", "모델", "크기", "ELO", "승", "패", "무", "승률"],
                datatype=["number", "str", "str", "number", "number", "number", "number", "str"],
            )

            gr.Markdown("### 투표 요약")
            with gr.Row():
                total_v = gr.Textbox(label="총 투표", interactive=False)
                a_wins_v = gr.Textbox(label="A 승", interactive=False)
                b_wins_v = gr.Textbox(label="B 승", interactive=False)
                ties_v = gr.Textbox(label="무승부", interactive=False)

            def refresh():
                lb = refresh_leaderboard()
                summary = get_vote_summary()
                return lb, *summary

            refresh_btn.click(refresh, outputs=[leaderboard, total_v, a_wins_v, b_wins_v, ties_v])

        # 모델 목록 탭
        with gr.Tab("모델 목록"):
            gr.Markdown("## 테스트 대상 모델")
            gr.Markdown(f"총 {len(MODELS)}개 모델")
            model_table = gr.Dataframe(
                headers=["모델 ID", "크기", "학습 방법", "설명", "Base Model"],
                value=[[k, v["size"], v["method"], v["desc"], v["base_model"]] for k, v in MODELS.items()],
            )

if __name__ == "__main__":
    demo.launch(server_name="0.0.0.0", server_port=7860)