Create model_trainer.py

learning_hub/model_trainer.py

@@ -0,0 +1,74 @@
+# learning_hub/model_trainer.py
+# (V32.0 - DORMANT: Model Retraining Logic)
+# ⚠️ This module is currently DISABLED by default.
+# It is kept for future manual triggering of model retraining.
+
+import os
+import asyncio
+import traceback
+import numpy as np
+import pandas as pd
+import xgboost as xgb
+from typing import List, Dict
+
+class ModelTrainer:
+    """
+    🏋️‍♂️ The Gym: Responsible for heavy lifting (Model Retraining).
+    Not active in the daily loop. Only triggered manually for regime updates.
+    """
+    def __init__(self, r2_service):
+        self.r2 = r2_service
+        self.local_model_path = "ml_models/layer2/Titan_XGB_V1.json"
+        print("💤 [ModelTrainer] Module loaded but dormant.")
+
+    async def train_model_manually(self, training_data: List[Dict]):
+        """
+        دالة للتدريب اليدوي عند الحاجة.
+        تتوقع قائمة من القواميس تحتوي على 'features' و 'label'.
+        """
+        print("🏋️‍♂️ [ModelTrainer] Manual training sequence started...")
+        try:
+            if not training_data:
+                print("❌ [Trainer] No data provided.")
+                return
+
+            # 1. Prepare Data
+            df = pd.DataFrame(training_data)
+            if 'features' not in df.columns or 'label' not in df.columns:
+                print("❌ [Trainer] Data missing features/label columns.")
+                return
+
+            X = np.array(df['features'].tolist())
+            y = np.array(df['label'].tolist())
+            dtrain = xgb.DMatrix(X, label=y)
+
+            # 2. Load Existing Model (Warm Start)
+            model = xgb.Booster()
+            if os.path.exists(self.local_model_path):
+                model.load_model(self.local_model_path)
+                print("   -> Loaded existing model for incremental training.")
+            
+            # 3. Train (Incremental)
+            params = {
+                'eta': 0.01,
+                'max_depth': 6,
+                'objective': 'binary:logistic',
+                'eval_metric': 'logloss'
+            }
+            
+            # تحديث النموذج بـ 50 جولة فقط
+            new_model = xgb.train(params, dtrain, num_boost_round=50, xgb_model=model)
+            
+            # 4. Save
+            new_model.save_model(self.local_model_path)
+            print(f"✅ [Trainer] Model updated and saved to {self.local_model_path}")
+            
+            # 5. Upload to R2 (Logic commented out until needed)
+            # await self.r2.upload_file_async(...)
+            
+            return "Training Success"
+
+        except Exception as e:
+            print(f"❌ [Trainer] Training failed: {e}")
+            traceback.print_exc()
+            return f"Error: {e}"