Spaces:
Sleeping
Sleeping
File size: 8,105 Bytes
1099afe bdedf43 1099afe |
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 |
# DEPENDENCIES
import sys
import threading
from enum import Enum
from typing import Any
from typing import Dict
from pathlib import Path
from typing import Optional
from dataclasses import field
from datetime import datetime
from dataclasses import dataclass
# Add parent directory to path for imports
sys.path.append(str(Path(__file__).parent.parent))
from utils.logger import log_info
from utils.logger import log_error
from utils.logger import ContractAnalyzerLogger
class ModelType(Enum):
"""
Enum for model types
"""
LEGAL_BERT = "legal-bert"
EMBEDDING = "embedding"
TOKENIZER = "tokenizer"
CLASSIFIER = "classifier"
class ModelStatus(Enum):
"""
Model loading status
"""
NOT_LOADED = "not_loaded"
LOADING = "loading"
LOADED = "loaded"
ERROR = "error"
@dataclass
class ModelInfo:
"""
Model metadata and state
"""
name : str
type : ModelType
status : ModelStatus = ModelStatus.NOT_LOADED
model : Optional[Any] = None
tokenizer : Optional[Any] = None
loaded_at : Optional[datetime] = None
error_message : Optional[str] = None
memory_size_mb : float = 0.0
access_count : int = 0
last_accessed : Optional[datetime] = None
metadata : Dict[str, Any] = field(default_factory = dict)
def mark_accessed(self):
"""
Update access statistics
"""
self.access_count += 1
self.last_accessed = datetime.now()
def get_age_seconds(self) -> float:
"""
Get seconds since last access
"""
if self.last_accessed:
return (datetime.now() - self.last_accessed).total_seconds()
return float('inf')
class ModelRegistry:
"""
Thread-safe singleton model registry : manages all loaded models with LRU eviction
"""
_instance = None
_lock = threading.Lock()
def __new__(cls):
if cls._instance is None:
with cls._lock:
if cls._instance is None:
cls._instance = super().__new__(cls)
cls._instance._initialized = False
return cls._instance
def __init__(self):
if self._initialized:
return
self._registry : Dict[ModelType, ModelInfo] = dict()
self._model_lock = threading.Lock()
self._max_models = 3 # LRU cache size
self._initialized = True
self.logger = ContractAnalyzerLogger.get_logger()
log_info("ModelRegistry initialized", max_models = self._max_models)
def register(self, model_type: ModelType, model_info: ModelInfo):
"""
Register a model (thread-safe)
"""
with self._model_lock:
self._registry[model_type] = model_info
self._enforce_cache_limit()
log_info(f"Model registered: {model_type.value}",
model_name = model_info.name,
status = model_info.status.value,
memory_mb = model_info.memory_size_mb,
)
def get(self, model_type: ModelType) -> Optional[ModelInfo]:
"""
Get model info (thread-safe)
"""
with self._model_lock:
info = self._registry.get(model_type)
if info:
info.mark_accessed()
log_info(f"Model accessed: {model_type.value}",
access_count = info.access_count,
status = info.status.value,
)
return info
def is_loaded(self, model_type: ModelType) -> bool:
"""
Check if model is loaded
"""
info = self.get(model_type)
return info is not None and (info.status == ModelStatus.LOADED)
def unload(self, model_type: ModelType):
"""
Unload a model from memory
"""
with self._model_lock:
if model_type in self._registry:
info = self._registry[model_type]
log_info(f"Unloading model: {model_type.value}",
memory_freed_mb = info.memory_size_mb,
access_count = info.access_count,
)
# Clear references to allow garbage collection
info.model = None
info.tokenizer = None
info.status = ModelStatus.NOT_LOADED
del self._registry[model_type]
def get_all_loaded(self) -> list[ModelInfo]:
"""
Get all loaded models
"""
with self._model_lock:
loaded = [info for info in self._registry.values() if (info.status == ModelStatus.LOADED)]
if loaded:
log_info(f"Retrieved {len(loaded)} loaded models", models = [info.name for info in loaded])
return loaded
def get_memory_usage(self) -> float:
"""
Get total memory usage in MB
"""
with self._model_lock:
total = sum(info.memory_size_mb for info in self._registry.values() if (info.status == ModelStatus.LOADED))
return total
def _enforce_cache_limit(self):
"""
Enforce LRU cache limit
"""
loaded_models = [(model_type, info) for model_type, info in self._registry.items() if (info.status == ModelStatus.LOADED)]
if (len(loaded_models) > self._max_models):
# Sort by last access time (oldest first)
loaded_models.sort(key = lambda x: x[1].get_age_seconds(), reverse = True)
# Unload oldest models
for model_type, info in loaded_models[self._max_models:]:
log_info(f"LRU eviction: {model_type.value}",
reason = "cache_limit_exceeded",
age_seconds = info.get_age_seconds(),
max_models = self._max_models,
)
self.unload(model_type)
def clear_all(self):
"""
Clear all models from registry
"""
with self._model_lock:
model_count = len(self._registry)
total_memory = self.get_memory_usage()
log_info("Clearing all models from registry",
models_cleared = model_count,
memory_freed_mb = total_memory,
)
for model_type in list(self._registry.keys()):
self.unload(model_type)
def get_stats(self) -> Dict[str, Any]:
"""
Get registry statistics
"""
with self._model_lock:
stats = {"total_models" : len(self._registry),
"loaded_models" : sum(1 for info in self._registry.values() if info.status == ModelStatus.LOADED),
"total_memory_mb" : self.get_memory_usage(),
"models" : {model_type.value: {"status" : info.status.value,
"access_count" : info.access_count,
"memory_mb" : info.memory_size_mb,
"last_accessed" : info.last_accessed.isoformat() if info.last_accessed else None,
}
for model_type, info in self._registry.items()
},
}
log_info("Registry stats retrieved", **stats)
return stats
|