updated app.py

app.py

@@ -6,7 +6,7 @@ import re
 
 # --- CONFIGURATION ---
 # REPLACE WITH YOUR USERNAME
-MODEL_ID = "justhariharan/Qwen2.5-Math-1.5B-Solver"
+MODEL_ID = "Hariharan123/Qwen2.5-Math-1.5B-Solver"
 
 print(f"⏳ Loading {MODEL_ID}... (CPU Mode)")
 tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
@@ -19,6 +19,7 @@ model = AutoModelForCausalLM.from_pretrained(
 # --- HELPER FUNCTIONS ---
 def extract_answer(text):
     """Extracts the number after #### or the last number found."""
+    if not text: return None
     if "####" in text:
         text = text.split("####")[-1]
     pattern = r"(-?[$0-9.,]{1,})"
@@ -35,13 +36,27 @@ You are a patient and friendly math teacher.
 
     # Add History (Short Term Memory - Last 1 Turn)
     history_context = ""
+    
+    # --- ROBUST HISTORY CHECK (The Fix) ---
     if len(history) > 0:
-        last_q, last_a = history[-1]
-        history_context = f"""
+        try:
+            # Get the last interaction
+            last_turn = history[-1]
+            
+            # Ensure it's a list/tuple
+            if isinstance(last_turn, (list, tuple)):
+                # Take only the first 2 elements (User, AI) and ignore extra metadata
+                last_q = last_turn[0]
+                last_a = last_turn[1]
+                
+                history_context = f"""
 <|im_start|>user
 {last_q}<|im_end|>
 <|im_start|>assistant
 {last_a}<|im_end|>"""
+        except Exception:
+            # If history format is weird, just ignore it and continue safely
+            pass
 
     # Current Input
     user_input = f"""
@@ -53,21 +68,24 @@ You are a patient and friendly math teacher.
 
 def solve_single(question, history, temperature=0.6):
     """Standard generation."""
-    prompt = format_prompt(question, history)
-    inputs = tokenizer(prompt, return_tensors="pt")
-    
-    with torch.no_grad():
-        outputs = model.generate(
-            **inputs, 
-            max_new_tokens=512,
-            temperature=temperature,
-            do_sample=True
-        )
-    
-    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    if "<|im_start|>assistant" in response:
-        return response.split("<|im_start|>assistant")[-1].strip()
-    return response
+    try:
+        prompt = format_prompt(question, history)
+        inputs = tokenizer(prompt, return_tensors="pt")
+        
+        with torch.no_grad():
+            outputs = model.generate(
+                **inputs, 
+                max_new_tokens=512,
+                temperature=temperature,
+                do_sample=True
+            )
+        
+        response = tokenizer.decode(outputs[0], skip_special_tokens=True)
+        if "<|im_start|>assistant" in response:
+            return response.split("<|im_start|>assistant")[-1].strip()
+        return response
+    except Exception as e:
+        return f"Error generating response: {str(e)}"
 
 def solve_majority_vote(question, history):
     """Smart Mode: Generates 3 answers and votes."""
@@ -101,7 +119,6 @@ def chat_logic(message, history, smart_mode):
     else:
         return solve_single(message, history)
 
-# --- UI SETUP ---
 # --- UI SETUP ---
 demo = gr.ChatInterface(
     fn=chat_logic,
@@ -112,18 +129,17 @@ demo = gr.ChatInterface(
     description="""
     <b>Portfolio Project:</b> A specialized math solver fine-tuned on GSM8K using LoRA.
     <br><br>
-    <b>Features:</b>
+    <b>⚠️ Performance Note:</b> This demo runs on <b>Free CPU Tier</b>.
     <ul>
-        <li><b>Teacher Persona:</b> Explains logic simply.</li>
-        <li><b>Smart Mode:</b> Uses 'Majority Voting' to boost accuracy from 70% → 82%.</li>
+        <li><b>First Request:</b> May take 1-2 mins (Cold Start).</li>
+        <li><b>Warm Requests:</b> ~10-20 seconds.</li>
+        <li><b>Smart Mode:</b> Runs 3x slower for higher accuracy.</li>
     </ul>
     """,
     examples=[
         ["If I have 30 candies and eat 12, then buy 5 more, how many do I have?", False],
-        ["It takes 5 machines 5 minutes to make 5 widgets. How long for 100 machines?", True],
-        ["Solve the integral solution for x + y + z = 15", True]
+        ["It takes 5 machines 5 minutes to make 5 widgets. How long for 100 machines?", True]
     ]
-    # theme="soft"  <-- THIS LINE WAS REMOVED
 )
 
 if __name__ == "__main__":