File size: 11,429 Bytes
ad05511
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
"""
安全模型加载器 - 从私有HF仓库加载模型
用于公开Space但保护模型文件
"""

import pickle
import pandas as pd
import numpy as np
import logging
import os
from pathlib import Path
from typing import Dict, Any, Optional
import warnings
warnings.filterwarnings('ignore')

# 安全模型加载 - 从私有HF仓库加载
try:
    from huggingface_hub import hf_hub_download
    HF_HUB_AVAILABLE = True
except ImportError:
    HF_HUB_AVAILABLE = False
    print("⚠️ huggingface_hub未安装,将使用本地模型文件")

# 配置日志
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

class SecureModelManager:
    """安全模型管理器 - 从私有HF仓库加载模型"""
    
    def __init__(self):
        """初始化安全模型管理器"""
        self.screening_models = {}
        self.advisory_models = {}
        self.thresholds = {}
        
        # HF私有仓库配置
        self.hf_repo_id = os.getenv("HF_MODEL_REPO", "YOUR_USERNAME/sarco-advisor-models")
        self.hf_token = os.getenv("HF_TOKEN")
        self.use_hf_models = HF_HUB_AVAILABLE and self.hf_token
        
        if self.use_hf_models:
            logger.info(f"🔒 使用HF私有仓库加载模型: {self.hf_repo_id}")
        else:
            logger.info("📁 回退到本地模型文件")
            # 导入原始模型管理器作为备用
            from .model_loader import ModelManager
            self.fallback_manager = ModelManager()
        
        # 加载所有模型
        self.load_all_models()
    
    def load_model_from_hf(self, model_path: str):
        """从HF私有仓库加载模型"""
        try:
            # 下载模型文件到临时位置
            local_path = hf_hub_download(
                repo_id=self.hf_repo_id,
                filename=model_path,
                token=self.hf_token,
                cache_dir="/tmp/hf_models"  # 临时缓存,不会被下载
            )
            
            # 加载模型
            with open(local_path, 'rb') as f:
                model = pickle.load(f)
            
            logger.info(f"✅ 从HF仓库加载模型: {model_path}")
            return model
            
        except Exception as e:
            logger.error(f"❌ HF模型加载失败 {model_path}: {str(e)}")
            return None
    
    def load_all_models(self):
        """加载所有模型"""
        if self.use_hf_models:
            self._load_models_from_hf()
        else:
            self._load_models_locally()
    
    def _load_models_from_hf(self):
        """从HF私有仓库加载所有模型"""
        logger.info("🔒 从HF私有仓库加载模型...")
        
        # 模型文件映射
        model_files = {
            # 筛查模型
            'sarcoI_screening': 'models/screening/sarcoI/randomforest_model.pkl',
            # 建议模型
            'sarcoI_advisory': 'models/advisory/sarcoI/CatBoost_model.pkl',
            'sarcoII_advisory': 'models/advisory/sarcoII/RandomForest_model.pkl'
        }
        
        # 阈值文件
        threshold_files = {
            'sarcoI_screening': 'models/screening/sarcoI/optimization_results.pkl',
            'sarcoII_screening': 'models/screening/sarcoII/optimization_results.pkl'
        }
        
        # 加载模型
        for model_name, model_path in model_files.items():
            model = self.load_model_from_hf(model_path)
            if model:
                if 'screening' in model_name:
                    model_type = model_name.replace('_screening', '')
                    self.screening_models[model_type] = model
                elif 'advisory' in model_name:
                    model_type = model_name.replace('_advisory', '')
                    self.advisory_models[model_type] = model
        
        # 加载阈值
        for threshold_name, threshold_path in threshold_files.items():
            threshold_data = self.load_model_from_hf(threshold_path)
            if threshold_data:
                model_type = threshold_name.replace('_screening', '')
                
                # 解析阈值数据
                if model_type == 'sarcoI':
                    if 'rf_best_threshold' in threshold_data:
                        self.thresholds[model_type] = {
                            'screening': threshold_data['rf_best_threshold'],
                            'advisory': 0.36  # 默认建议模型阈值
                        }
                elif model_type == 'sarcoII':
                    if 'catboost_best_threshold' in threshold_data:
                        self.thresholds[model_type] = {
                            'screening': threshold_data['catboost_best_threshold'],
                            'advisory': 0.52  # 默认建议模型阈值
                        }
        
        logger.info("✅ HF模型加载完成")
    
    def _load_models_locally(self):
        """回退到本地模型加载"""
        logger.info("📁 使用本地模型文件...")
        
        if hasattr(self, 'fallback_manager'):
            self.fallback_manager.load_all_models()
            
            # 复制模型和阈值
            self.screening_models = self.fallback_manager.screening_models
            self.advisory_models = self.fallback_manager.advisory_models
            self.thresholds = self.fallback_manager.thresholds
            
            logger.info("✅ 本地模型加载完成")
    
    def predict_screening(self, user_data: Dict, model_type: str) -> Dict:
        """筛查预测"""
        if hasattr(self, 'fallback_manager') and not self.use_hf_models:
            return self.fallback_manager.predict_screening(user_data, model_type)
        
        # HF模型预测逻辑
        if model_type not in self.screening_models:
            raise ValueError(f"筛查模型 {model_type} 未找到")
        
        model = self.screening_models[model_type]
        threshold = self.thresholds.get(model_type, {}).get('screening', 0.5)
        
        # 准备特征数据
        if model_type == 'sarcoI':
            features = ['age_years', 'WWI', 'body_mass_index']
        else:
            features = ['age_years', 'race_ethnicity', 'WWI', 'body_mass_index']
        
        X = np.array([[user_data[f] for f in features]])
        
        # 预测
        probability = model.predict_proba(X)[0][1]
        risk_level = 'high' if probability >= threshold else 'low'
        
        return {
            'probability': probability,
            'risk_level': risk_level,
            'threshold': threshold,
            'model_type': model_type
        }
    
    def predict_advisory(self, user_data: Dict, model_type: str) -> Dict:
        """建议预测"""
        if hasattr(self, 'fallback_manager') and not self.use_hf_models:
            return self.fallback_manager.predict_advisory(user_data, model_type)
        
        # HF模型预测逻辑
        if model_type not in self.advisory_models:
            raise ValueError(f"建议模型 {model_type} 未找到")
        
        model = self.advisory_models[model_type]
        threshold = self.thresholds.get(model_type, {}).get('advisory', 0.5)
        
        # 准备特征数据(简化版本)
        if model_type == 'sarcoI':
            features = ['body_mass_index', 'race_ethnicity', 'WWI', 'age_years', 
                       'Activity_Sedentary_Ratio', 'Total_Moderate_Minutes_week', 'Vigorous_MET_Ratio']
        else:
            features = ['body_mass_index', 'race_ethnicity', 'age_years', 'Activity_Sedentary_Ratio',
                       'Activity_Diversity_Index', 'WWI', 'Vigorous_MET_Ratio', 'sedentary_minutes']
        
        # 检查特征是否存在
        available_features = []
        for f in features:
            if f in user_data:
                available_features.append(user_data[f])
            else:
                available_features.append(0.0)  # 默认值
        
        X = np.array([available_features])
        
        # 预测
        probability = model.predict_proba(X)[0][1]
        risk_level = 'high' if probability >= threshold else 'low'
        
        return {
            'probability': probability,
            'risk_level': risk_level,
            'threshold': threshold,
            'model_type': model_type
        }
    
    def get_comprehensive_risk(self, sarcoI_screening_result: Dict, sarcoI_advisory_result: Dict = None,
                              sarcoII_screening_result: Dict = None, sarcoII_advisory_result: Dict = None) -> Dict:
        """综合风险评估"""
        if hasattr(self, 'fallback_manager') and not self.use_hf_models:
            return self.fallback_manager.get_comprehensive_risk(
                sarcoI_screening_result, sarcoI_advisory_result,
                sarcoII_screening_result, sarcoII_advisory_result
            )
        
        # 使用与原始模型管理器相同的逻辑
        results = {}

        # SarcoI 综合风险判定
        if sarcoI_screening_result:
            P_recall_I = sarcoI_screening_result['probability']
            P_precision_I = sarcoI_advisory_result['probability'] if sarcoI_advisory_result else 0.0
            
            sarcoI_advisory_threshold = self.thresholds.get('sarcoI', {}).get('advisory', 0.36)
            sarcoI_screening_threshold = self.thresholds.get('sarcoI', {}).get('screening', 0.23)

            if P_precision_I >= sarcoI_advisory_threshold:
                sarcoI_comprehensive_risk = "high"
                sarcoI_risk_reason = "advisory_model_high_risk"
            elif P_recall_I >= sarcoI_screening_threshold:
                sarcoI_comprehensive_risk = "medium"
                sarcoI_risk_reason = "screening_model_risk"
            else:
                sarcoI_comprehensive_risk = "low"
                sarcoI_risk_reason = "both_models_low_risk"

            results['sarcoI'] = {
                'comprehensive_risk': sarcoI_comprehensive_risk,
                'screening_probability': P_recall_I,
                'advisory_probability': P_precision_I,
                'risk_reason': sarcoI_risk_reason
            }

        # SarcoII 综合风险判定
        if sarcoII_screening_result:
            P_recall_II = sarcoII_screening_result['probability']
            P_precision_II = sarcoII_advisory_result['probability'] if sarcoII_advisory_result else 0.0
            
            sarcoII_advisory_threshold = self.thresholds.get('sarcoII', {}).get('advisory', 0.52)
            sarcoII_screening_threshold = self.thresholds.get('sarcoII', {}).get('screening', 0.15)

            if P_precision_II >= sarcoII_advisory_threshold:
                sarcoII_comprehensive_risk = "high"
                sarcoII_risk_reason = "advisory_model_high_risk"
            elif P_recall_II >= sarcoII_screening_threshold:
                sarcoII_comprehensive_risk = "medium"
                sarcoII_risk_reason = "screening_model_risk"
            else:
                sarcoII_comprehensive_risk = "low"
                sarcoII_risk_reason = "both_models_low_risk"

            results['sarcoII'] = {
                'comprehensive_risk': sarcoII_comprehensive_risk,
                'screening_probability': P_recall_II,
                'advisory_probability': P_precision_II,
                'risk_reason': sarcoII_risk_reason
            }

        return results