Create models.py

models.py

@@ -1,29 +1,4 @@
-def predict_intent(self, text):
-        """Predict intent using multilabel classification"""
-        if not text.strip():
-            return "neutral", 0.5
-        
-        # Check if intent model is available
-        if "intent" not in self.models:
-            logger.warning("Intent model not available, returning neutral intent")
-            return "neutral", 0.5
-            
-        try:
-            inputs = self._prepare_inputs("intent", text)
-            
-            with torch.no_grad():
-                outputs = self.models["intent"](**inputs)
-                
-                if self.models["intent"].is_multilabel:
-                    # For multilabel, use sigmoid and get all scores
-                    probs = torch.sigmoid(outputs.logits).cpu().numpy()[0]
-                    logger.debug(f"Intent model raw probabilities: {probs}")
-                    
-                    # Get intent labels - you may need to adjust these based on your actual model
-                    intent_labels = ["neutral", "manipulative", "supportive", "controlling", "threatening", "gaslighting"]
-                    
-                    # Debug: print all scores
-                   
+import torch
 import torch.nn as nn
 import logging
 import os
@@ -51,10 +26,6 @@ class MultiLabelIntentClassifier(nn.Module):
         logits = self.classifier(pooled_output)
         return logits
 
-# Set up logging
-logging.basicConfig(level=logging.INFO)
-logger = logging.getLogger(__name__)
-
 class ModelManager:
     def __init__(self, device=None):
         """Initialize model manager with device detection"""
@@ -270,7 +241,6 @@ class ModelManager:
         """Load emotion pipeline with retry logic"""
         for attempt in range(max_retries):
             try:
-                from transformers import pipeline
                 logger.info(f"Loading emotion pipeline (attempt {attempt+1}/{max_retries})")
                 
                 self.emotion_pipeline = pipeline(
@@ -289,9 +259,8 @@ class ModelManager:
                 logger.error(f"Error loading emotion pipeline (attempt {attempt+1}): {e}")
                 time.sleep(2)  # Wait before retry
         
-        if not emotion_success:
-            logger.error("Failed to load emotion pipeline after all retries")
-            return False
+        logger.error("Failed to load emotion pipeline after all retries")
+        return False
     
     def predict_fallacy(self, text):
         """Predict logical fallacy using FallacyFinder model"""