File size: 9,943 Bytes
edf1149 |
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 |
# DEPENDENCIES
import gc
import torch
import threading
from typing import Any
from typing import Dict
from typing import List
from loguru import logger
from typing import Optional
from datetime import datetime
from dataclasses import dataclass
from config.model_config import ModelConfig
from config.model_config import MODEL_REGISTRY
from config.model_config import get_model_config
@dataclass
class ModelUsageStats:
"""
Lightweight model usage statistics
"""
model_name : str
load_count : int
last_used : datetime
total_usage_time_seconds : float
avg_usage_time_seconds : float
def to_dict(self) -> Dict[str, Any]:
"""
Convert to dictionary
"""
return {"model_name" : self.model_name,
"load_count" : self.load_count,
"last_used" : self.last_used.isoformat(),
"total_usage_time_seconds" : round(self.total_usage_time_seconds, 2),
"avg_usage_time_seconds" : round(self.avg_usage_time_seconds, 2),
}
class ModelRegistry:
"""
Model registry module for tracking model usage statistics and performance metrics
Complements ModelManager by adding:
- Usage analytics
- Performance monitoring
- Model dependency tracking
- Health checks (without duplicating ModelManager functionality)
"""
def __init__(self):
self.usage_stats : Dict[str, ModelUsageStats] = dict()
self.dependency_graph : Dict[str, List[str]] = dict()
self.performance_metrics : Dict[str, Dict[str, float]] = dict()
self.lock = threading.RLock()
# Initialize from MODEL_REGISTRY
self._initialize_registry()
logger.info("ModelRegistry initialized for usage tracking")
def _initialize_registry(self):
"""
Initialize registry with all known models
"""
for model_name in MODEL_REGISTRY.keys():
self.usage_stats[model_name] = ModelUsageStats(model_name = model_name,
load_count = 0,
last_used = datetime.now(),
total_usage_time_seconds = 0.0,
avg_usage_time_seconds = 0.0,
)
def record_model_usage(self, model_name: str, usage_time_seconds: float = 0.0):
"""
Record that a model was used
Arguments:
----------
model_name { str } : Name of the model used
usage_time_seconds { float } : How long the model was used (if available)
"""
with self.lock:
if model_name not in self.usage_stats:
# Auto-register unknown models
self.usage_stats[model_name] = ModelUsageStats(model_name = model_name,
load_count = 0,
last_used = datetime.now(),
total_usage_time_seconds = 0.0,
avg_usage_time_seconds = 0.0,
)
stats = self.usage_stats[model_name]
stats.load_count += 1
stats.last_used = datetime.now()
if (usage_time_seconds > 0):
stats.total_usage_time_seconds += usage_time_seconds
stats.avg_usage_time_seconds = stats.total_usage_time_seconds / stats.load_count
logger.debug(f"Recorded usage for {model_name} (count: {stats.load_count})")
def get_usage_stats(self, model_name: str) -> Optional[ModelUsageStats]:
"""
Get usage statistics for a model
"""
with self.lock:
return self.usage_stats.get(model_name)
def get_most_used_models(self, top_k: int = 5) -> List[ModelUsageStats]:
"""
Get most frequently used models
"""
with self.lock:
sorted_models = sorted(self.usage_stats.values(),
key = lambda x: x.load_count,
reverse = True,
)
return sorted_models[:top_k]
def record_performance_metric(self, model_name: str, metric_name: str, value: float):
"""
Record performance metrics for a model
Arguments:
----------
model_name { str } : Name of the model
metric_name { float } : Name of the metric (e.g., "inference_time_ms", "memory_peak_mb")
value { str } : Metric value
"""
with self.lock:
if model_name not in self.performance_metrics:
self.performance_metrics[model_name] = {}
self.performance_metrics[model_name][metric_name] = value
def get_performance_metrics(self, model_name: str) -> Dict[str, float]:
"""
Get performance metrics for a model
"""
with self.lock:
return self.performance_metrics.get(model_name, {})
def add_dependency(self, model_name: str, depends_on: List[str]):
"""
Add dependency information for a model
Arguments:
----------
model_name { str } : The model that has dependencies
depends_on { list } : List of model names this model depends on
"""
with self.lock:
self.dependency_graph[model_name] = depends_on
def get_dependencies(self, model_name: str) -> List[str]:
"""
Get dependencies for a model
"""
with self.lock:
return self.dependency_graph.get(model_name, [])
def get_dependent_models(self, model_name: str) -> List[str]:
"""
Get models that depend on the specified model
"""
with self.lock:
dependents = []
for user_model, dependencies in self.dependency_graph.items():
if model_name in dependencies:
dependents.append(user_model)
return dependents
def generate_usage_report(self) -> Dict[str, Any]:
"""
Generate a comprehensive usage report
"""
with self.lock:
total_usage = sum(stats.load_count for stats in self.usage_stats.values())
active_models = [name for name, stats in self.usage_stats.items() if stats.load_count > 0]
return {"timestamp" : datetime.now().isoformat(),
"summary" : {"total_models_tracked" : len(self.usage_stats),
"active_models" : len(active_models),
"total_usage_count" : total_usage,
},
"most_used_models" : [stats.to_dict() for stats in self.get_most_used_models(top_k = 10)],
"performance_metrics" : {model: metrics for model, metrics in self.performance_metrics.items()},
"dependency_graph" : self.dependency_graph
}
def reset_usage_stats(self, model_name: Optional[str] = None):
"""
Reset usage statistics for a model or all models
Arguments:
----------
model_name { str } : Specific model to reset, or None for all models
"""
with self.lock:
if model_name:
if model_name in self.usage_stats:
self.usage_stats[model_name] = ModelUsageStats(model_name = model_name,
load_count = 0,
last_used = datetime.now(),
total_usage_time_seconds = 0.0,
avg_usage_time_seconds = 0.0,
)
logger.info(f"Reset usage stats for {model_name}")
else:
self._initialize_registry()
logger.info("Reset usage stats for all models")
def cleanup(self):
"""
Clean up resources
"""
with self.lock:
self.usage_stats.clear()
self.performance_metrics.clear()
self.dependency_graph.clear()
logger.info("ModelRegistry cleanup completed")
# Singleton instance
_model_registry_instance: Optional[ModelRegistry] = None
_registry_lock = threading.Lock()
def get_model_registry() -> ModelRegistry:
"""
Get singleton ModelRegistry instance
"""
global _model_registry_instance
if _model_registry_instance is None:
with _registry_lock:
if _model_registry_instance is None:
_model_registry_instance = ModelRegistry()
return _model_registry_instance
# Export
__all__ = ["ModelRegistry",
"ModelUsageStats",
"get_model_registry"
] |