Update inference.py

inference.py

@@ -18,7 +18,7 @@ tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased")
 model = None
 last_mod_time = 0
 
-# 🔁 Reload model if changed on disk
+# 🔁 Reload model if updated on disk
 def load_model():
     global model, last_mod_time
     try:
@@ -34,18 +34,18 @@ def load_model():
         model = None
     return model
 
-# 🧠 Evo logic
+# 🧠 Evo inference logic
 def get_evo_response(query, options, user_context=""):
     model = load_model()
     if model is None:
         return "Error", 0.0, "Model failed to load", ""
 
-    # Retrieve web search + optional user context
+    # Context = web + user
     context_texts = web_search(query) + ([user_context] if user_context else [])
     context_str = "\n".join(context_texts)
     input_pairs = [f"{query} [SEP] {opt} [CTX] {context_str}" for opt in options]
 
-    # Encode and score each option
+    # Score each option
     scores = []
     for pair in input_pairs:
         encoded = tokenizer(pair, return_tensors="pt", padding="max_length", truncation=True, max_length=128)
@@ -56,13 +56,13 @@ def get_evo_response(query, options, user_context=""):
 
     best_idx = int(scores[1] > scores[0])
     return (
-        options[best_idx],                      # ✅ Evo's answer
+        options[best_idx],                      # ✅ Evo's best answer
         max(scores),                            # ✅ Confidence
         f"{options[0]}: {scores[0]:.3f} vs {options[1]}: {scores[1]:.3f}",  # ✅ Reasoning trace
-        context_str                             # ✅ Context used
+        context_str                             # ✅ Used context
     )
 
-# 🔄 GPT backup response
+# 🔄 GPT-3.5 response
 def get_gpt_response(query, user_context=""):
     try:
         context_block = f"\n\nContext:\n{user_context}" if user_context else ""
@@ -75,7 +75,7 @@ def get_gpt_response(query, user_context=""):
     except Exception as e:
         return f"⚠️ GPT error:\n\n{str(e)}"
 
-# 🎯 Used by app.py to display Evo live output
+# 🎯 For EvoRAG app UI
 def evo_chat_predict(history, query, options):
     context = "\n".join(history[-6:]) if history else ""
     evo_ans, evo_score, evo_reason, evo_ctx = get_evo_response(query, options, context)
@@ -86,15 +86,17 @@ def evo_chat_predict(history, query, options):
         "context_used": evo_ctx
     }
 
-# 📊 Returns current Evo architecture stats (for UI display)
+# 📊 Evo architecture stats
 def get_model_config():
     return {
         "num_layers": 6,
         "num_heads": 8,
         "ffn_dim": 1024,
-        "memory_enabled": True
+        "memory_enabled": True,
+        "param_count": sum(p.numel() for p in model.parameters() if p.requires_grad) if model else "N/A"
     }
 
+# 💻 Hardware and system stats
 def get_system_stats():
     gpu_info = torch.cuda.get_device_properties(0) if torch.cuda.is_available() else None
     memory = psutil.virtual_memory()