Update app.py

app.py

@@ -3,40 +3,33 @@ import gradio as gr
 import torch
 import time
 
-# Load model and tokenizer
 model = GPT2LMHeadModel.from_pretrained("GBhaveshKumar/ConvoAI")
 tokenizer = AutoTokenizer.from_pretrained("GBhaveshKumar/ConvoAI")
 
-# Use CPU
-device = torch.device("cpu")
+device = torch.device("cpu") 
 model = model.to(device)
 
-# Response generator
 def respond(message, history):
-    # Simple 1-turn prompt (faster for CPU)
     context = f"A: {message}\nB:"
-    inputs = tokenizer(context, return_tensors="pt", max_length=96, truncation=True)
+    inputs = tokenizer(context, return_tensors="pt", truncation=True, max_length=128)
     inputs = {k: v.to(device) for k, v in inputs.items()}
 
-    # Measure response time
     start_time = time.time()
     outputs = model.generate(
         **inputs,
-        max_new_tokens=24,  # Smaller = faster
+        max_new_tokens=32,
         do_sample=False,
         pad_token_id=tokenizer.eos_token_id
     )
     duration = time.time() - start_time
     print(f"⏱️ Response time: {duration:.2f} seconds")
 
-    # Decode only new tokens
-    generated_text = tokenizer.decode(outputs[0])
-    reply = generated_text.split("B:")[-1].strip().split("\n")[0]
-    
+    decoded = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    reply = decoded[len(context):].strip().split("\n")[0]
+
     history.append((message, reply))
     return history
 
-# Gradio UI
 with gr.Blocks() as demo:
     gr.Markdown("<h1 style='text-align: center; color: #2e8b57;'>🤖 ConvoAI</h1>")
     gr.Markdown("<p style='text-align: center; font-size: 18px;'>Conversational AI trained using the DailyDialog dataset. Engage in meaningful and natural conversations!</p>")