Update app.py

app.py

@@ -13,24 +13,36 @@ device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 model.to(device)
 
 # Function to generate response
-def generate_response(prompt, model, tokenizer, max_length=100, num_beams=5, temperature=0.5, top_p=0.9, repetition_penalty=4.0):
-    # Prepare the prompt
-    prompt = f"User: {prompt}\nAssistant:"
+def generate_response(message, history, system_message, max_tokens, temperature, top_p):
+    # Prepare the conversation history
+    messages = [{"role": "system", "content": system_message}]
+    
+    for user_msg, bot_msg in history:
+        if user_msg:
+            messages.append({"role": "user", "content": user_msg})
+        if bot_msg:
+            messages.append({"role": "assistant", "content": bot_msg})
+    
+    messages.append({"role": "user", "content": message})
+    
+    # Tokenize and prepare the input
+    prompt = "\n".join([f"{msg['role'].capitalize()}: {msg['content']}" for msg in messages])
     inputs = tokenizer(prompt, return_tensors='pt', padding=True, truncation=True, max_length=512).to(device)
+    
+    # Generate the response
     outputs = model.generate(
         inputs['input_ids'],
-        max_length=max_length,
+        max_length=max_tokens,
         num_return_sequences=1,
         pad_token_id=tokenizer.eos_token_id,
-        num_beams=num_beams,
         temperature=temperature,
         top_p=top_p,
-        repetition_penalty=repetition_penalty,
-        early_stopping=True
+        early_stopping=True,
+        do_sample=True  # Enable sampling
     )
     response = tokenizer.decode(outputs[0], skip_special_tokens=True)
     
-    # Post-processing to clean up the response
+    # Clean up the response
     response = response.split("Assistant:")[-1].strip()
     response_lines = response.split('\n')
     clean_response = []
@@ -39,13 +51,12 @@ def generate_response(prompt, model, tokenizer, max_length=100, num_beams=5, tem
             clean_response.append(line)
     response = ' '.join(clean_response)
     
-    return 
-    
-    return [(prompt, response.strip())]
+    return [(message, response)]
 
 # Create the Gradio chat interface
 demo = gr.ChatInterface(
     fn=generate_response,
+  
     title="Chatbot",
     description="Ask anything to the chatbot."
 )