Hugging Face's logo

Spaces:
YoMioAI
/
Roleplay_LLM_Arena
Runtime error

App Files Community
Roleplay_LLM_Arena
File size: 3,868 Bytes 
4182c57
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
87ecac3
 
4182c57
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e83906e
 
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
 
import time
from pymongo import MongoClient
import pandas as pd
import math
import os

# MongoDB连接配置
client = MongoClient(os.getenv("client_link"))
db = client.get_database('roleplay')  
collection = db.get_collection('model_stats')  

def update_model_stats(model1_name, model2_name, winner, turn, anony, language):
    # 获取当前时间戳
    tstamp = time.time()
    
    # 插入数据到MongoDB
    collection.insert_one({
        "Model 1": model1_name,
        "Model 2": model2_name,
        "Winner": winner,
        "Turn": turn,
        "Anony": anony,
        "Language": language,
        "tstamp": tstamp
    })

def calculate_elo(winner_elo, loser_elo, k=30, outcome=1):
    """
    winner_elo: Elo score of the winner before the game
    loser_elo: Elo score of the loser before the game
    k: K-factor in Elo calculation
    outcome: 1 if winner won, 0.5 if tie, 0 if loser won (inverted)
    """
    expected_win = 1 / (1 + math.pow(10, (loser_elo - winner_elo) / 400))
    new_winner_elo = winner_elo + k * (outcome - expected_win)
    return new_winner_elo

def load_dataframe():
    # 从MongoDB读取数据
    cursor = collection.find({})
    
    # 将游标中的数据转换为DataFrame
    data = pd.DataFrame(list(cursor))

    # 创建模型名称的唯一列表
    models = pd.unique(data[['Model 1', 'Model 2']].values.ravel('K'))

    # 初始化结果字典
    results = {'模型名称': [], '参赛次数': [], '胜利次数': [], 'ELO': []}
    elo_dict = {model: 1000 for model in models}  # 初始化ELO分数为1000

    for _, row in data.iterrows():
        model1 = row['Model 1']
        model2 = row['Model 2']
        winner = row['Winner']

        if winner == 'Model 1':
            elo_dict[model1] = calculate_elo(elo_dict[model1], elo_dict[model2], outcome=1)
            elo_dict[model2] = calculate_elo(elo_dict[model2], elo_dict[model1], outcome=0)
        elif winner == 'Model 2':
            elo_dict[model2] = calculate_elo(elo_dict[model2], elo_dict[model1], outcome=1)
            elo_dict[model1] = calculate_elo(elo_dict[model1], elo_dict[model2], outcome=0)
        elif winner == 'tie':
            elo_dict[model1] = calculate_elo(elo_dict[model1], elo_dict[model2], outcome=0.8)
            elo_dict[model2] = calculate_elo(elo_dict[model2], elo_dict[model1], outcome=0.8)
        elif winner == 'bothbad':
            elo_dict[model1] = calculate_elo(elo_dict[model1], elo_dict[model2], outcome=0.05)
            elo_dict[model2] = calculate_elo(elo_dict[model2], elo_dict[model1], outcome=0.05)

    for model in models:
        count = data['Model 1'].value_counts().get(model, 0) + data['Model 2'].value_counts().get(model, 0)
        win_count = 0
        win_count += len(data[(data['Winner'] == 'Model 1') & (data['Model 1'] == model)])
        win_count += len(data[(data['Winner'] == 'Model 2') & (data['Model 2'] == model)])
        win_count += len(data[(data['Winner'] == 'tie') & ((data['Model 1'] == model) | (data['Model 2'] == model))])
        results['模型名称'].append(model)
        results['参赛次数'].append(count)
        results['胜利次数'].append(win_count)
        results['ELO'].append(round(elo_dict[model]))

    # 将结果字典转换为DataFrame
    result_df = pd.DataFrame(results)

    # 计算胜率并排序
    result_df["模型胜率"] = (result_df['胜利次数'] / result_df['参赛次数']) * 100
    result_df = result_df.sort_values(by="模型胜率", ascending=False)
    result_df["模型胜率"] = result_df["模型胜率"].map("{:.2f}%".format)
    
    return result_df

def change_name(old,new):
    collection.update_many(
        { "Model 1": old },
        { "$set": { "Model 1": new } }
    )

    # 更新 Model 2 字段
    collection.update_many(
        { "Model 2": old },
        { "$set": { "Model 2": new } }
    )