Spaces:
Running
Running
File size: 8,977 Bytes
09801ca | 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 | """
Model Deployer — One-Click ML Model Deployment
================================================
Manages deployment lifecycle: deploy, predict, undeploy, status.
Uses a JSON file registry for deployments and caches loaded engines in memory.
"""
import json
import uuid
import time
import datetime
import logging
from pathlib import Path
from typing import Dict, Any, Optional, List
from ml.model_persistence import get_model_persistence_manager
from ml.automl_engine import ProductionMLEngine
logger = logging.getLogger(__name__)
class ModelDeployer:
"""Manages one-click deployment of trained ML models"""
def __init__(self, storage_dir: str = "storage/deployments"):
self.storage_dir = Path(storage_dir)
self.storage_dir.mkdir(parents=True, exist_ok=True)
self.registry_path = self.storage_dir / "registry.json"
# Load registry
self.registry = self._load_registry()
# Cache for loaded engines
self._engines: Dict[str, ProductionMLEngine] = {}
# =========================================================================
# Registry I/O
# =========================================================================
def _load_registry(self) -> Dict[str, Any]:
if self.registry_path.exists():
try:
with open(self.registry_path, 'r') as f:
return json.load(f)
except Exception:
return {}
return {}
def _save_registry(self):
with open(self.registry_path, 'w') as f:
json.dump(self.registry, f, indent=2)
# =========================================================================
# Engine Loading (single source of truth)
# =========================================================================
def _load_engine(self, state: Dict[str, Any]) -> ProductionMLEngine:
"""
Hydrate a ProductionMLEngine from a persisted model state dict.
This is the ONLY place engine fields are assigned — no duplication.
"""
engine = ProductionMLEngine()
engine.model = state.get('model')
engine.model_name = state.get('model_name', 'Unknown Model')
engine.task_type = state.get('task_type', 'unknown')
engine.task_type_simple = (
'classification' if 'classification' in engine.task_type else 'regression'
)
engine.feature_columns = state.get('feature_columns', [])
engine.target_column = state.get('target_column', '')
engine.label_encoders = state.get('label_encoders', {})
engine.scaler = state.get('scaler')
engine.numeric_cols = state.get('numeric_cols', [])
engine.categorical_cols = state.get('categorical_cols', [])
return engine
def _get_or_load_engine(self, deploy_id: str) -> ProductionMLEngine:
"""Get engine from cache, or load from persistence."""
if deploy_id in self._engines:
return self._engines[deploy_id]
deployment = self.registry.get(deploy_id)
if not deployment:
raise ValueError(f"Deployment {deploy_id} not found")
pm = get_model_persistence_manager()
state = pm.load_model(deployment["user_id"], version=deployment.get("version"))
if not state:
raise ValueError("Underlying model data lost — cannot load engine")
engine = self._load_engine(state)
self._engines[deploy_id] = engine
return engine
# =========================================================================
# Deploy / Undeploy
# =========================================================================
def deploy_model(self, user_id: str, version: Optional[int] = None) -> Dict[str, Any]:
"""Deploy a user's model and return deployment metadata."""
pm = get_model_persistence_manager()
# Verify model exists
state = pm.load_model(user_id, version=version)
if not state:
raise ValueError(
f"No model found for user {user_id}"
+ (f" (version {version})" if version else "")
)
model_name = state.get('model_name', 'Unknown Model')
task_type = state.get('task_type', 'unknown')
# Generate deploy ID and API key
deploy_id = f"deploy_{uuid.uuid4().hex[:8]}"
api_key = f"dv_{uuid.uuid4().hex}"
deployment = {
"deploy_id": deploy_id,
"user_id": user_id,
"version": version,
"model_name": model_name,
"task_type": task_type,
"created_at": datetime.datetime.now().isoformat(),
"status": "active",
"api_key": api_key,
"endpoint": f"/api/v1/deploy/predict/{deploy_id}",
"request_count": 0,
}
self.registry[deploy_id] = deployment
self._save_registry()
# Pre-load engine into cache
self._engines[deploy_id] = self._load_engine(state)
logger.info(f"Deployed model '{model_name}' as {deploy_id} for user {user_id}")
return deployment
def undeploy(self, deploy_id: str, user_id: Optional[str] = None) -> Dict[str, Any]:
"""Deactivate a deployment. Optionally verify ownership via user_id."""
if deploy_id not in self.registry:
raise ValueError(f"Deployment {deploy_id} not found")
deployment = self.registry[deploy_id]
if user_id and deployment["user_id"] != user_id:
raise ValueError("You do not own this deployment")
deployment["status"] = "inactive"
deployment["deactivated_at"] = datetime.datetime.now().isoformat()
self._save_registry()
# Evict from engine cache
self._engines.pop(deploy_id, None)
logger.info(f"Undeployed {deploy_id}")
return {"deploy_id": deploy_id, "status": "inactive"}
# =========================================================================
# Query
# =========================================================================
def list_deployments(self, user_id: str) -> List[Dict[str, Any]]:
return [d for d in self.registry.values() if d["user_id"] == user_id]
def get_deployment_status(self, deploy_id: str) -> Dict[str, Any]:
"""Return health / metadata for a single deployment."""
if deploy_id not in self.registry:
raise ValueError(f"Deployment {deploy_id} not found")
deployment = self.registry[deploy_id]
# Check if engine is cached (warm) or needs loading (cold)
engine_loaded = deploy_id in self._engines
return {
"deploy_id": deploy_id,
"status": deployment.get("status", "unknown"),
"model_name": deployment.get("model_name"),
"task_type": deployment.get("task_type"),
"created_at": deployment.get("created_at"),
"request_count": deployment.get("request_count", 0),
"engine_loaded": engine_loaded,
"endpoint": deployment.get("endpoint"),
}
# =========================================================================
# Inference
# =========================================================================
def predict(self, deploy_id: str, data: Dict[str, Any], api_key: str = None) -> Dict[str, Any]:
"""Run inference against a deployed model."""
if deploy_id not in self.registry:
raise ValueError(f"Deployment {deploy_id} not found")
deployment = self.registry[deploy_id]
if deployment.get("status") != "active":
raise ValueError(f"Deployment {deploy_id} is not active (status: {deployment.get('status')})")
# Simple API key auth
if api_key and deployment["api_key"] != api_key:
raise ValueError("Invalid API key")
engine = self._get_or_load_engine(deploy_id)
start_time = time.time()
try:
result = engine.predict(data)
duration_ms = (time.time() - start_time) * 1000
# Increment request counter
deployment["request_count"] = deployment.get("request_count", 0) + 1
self._save_registry()
# Log telemetry
try:
from ml.model_monitor import ModelMonitor
ModelMonitor.log_inference(deploy_id, data, duration_ms)
except Exception as e:
logger.debug(f"Telemetry logging skipped: {e}")
return result
except Exception as e:
raise ValueError(f"Prediction error: {str(e)}")
# =============================================================================
# Singleton
# =============================================================================
_deployer = None
def get_model_deployer() -> ModelDeployer:
global _deployer
if _deployer is None:
_deployer = ModelDeployer()
return _deployer
|