Update app.py

app.py

@@ -1,48 +1,37 @@
 
-import os
 import gradio as gr
-from transformers import AutoModelForCausalLM, AutoTokenizer
 import torch
+from transformers import GPT2Tokenizer, GPT2LMHeadModel
 
-# Load your model and tokenizer from Hugging Face
-model_name = 'redael/model_udc'
-tokenizer = AutoTokenizer.from_pretrained(model_name)
-model = AutoModelForCausalLM.from_pretrained(model_name)
+# Load the custom model and tokenizer
+model_path =  'redael/model_udc'
+tokenizer = GPT2Tokenizer.from_pretrained(model_path)
+model = GPT2LMHeadModel.from_pretrained(model_path)
 
+# Check if CUDA is available and use GPU if possible, enable FP16 precision
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 model.to(device)
+if device.type == 'cuda':
+    model = model.half()  # Use FP16 precision
 
-# Function to generate response
-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])
+def generate_response(prompt, model, tokenizer, max_length=100, num_beams=1, temperature=0.7, top_p=0.9, repetition_penalty=1.0):
+    # Prepare the prompt
+    prompt = f"User: {prompt}\nAssistant:"
     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_tokens,
+        max_length=max_length,
         num_return_sequences=1,
         pad_token_id=tokenizer.eos_token_id,
+        num_beams=num_beams,  # Use a lower number of beams
         temperature=temperature,
         top_p=top_p,
-        early_stopping=True,
-        do_sample=True  # Enable sampling
+        repetition_penalty=repetition_penalty,
+        early_stopping=True
     )
     response = tokenizer.decode(outputs[0], skip_special_tokens=True)
     
-    # Clean up the response
+    # Post-processing to clean up the response
     response = response.split("Assistant:")[-1].strip()
     response_lines = response.split('\n')
     clean_response = []
@@ -50,15 +39,28 @@ def generate_response(message, history, system_message, max_tokens, temperature,
         if "User:" not in line and "Assistant:" not in line:
             clean_response.append(line)
     response = ' '.join(clean_response)
+    return response.strip()
+
+def respond(message, history: list[tuple[str, str]]):
+    # Prepare the prompt from the history and the new message
+    system_message = "You are a friendly chatbot."
+    conversation = system_message + "\n"
+    for user_message, assistant_response in history:
+        conversation += f"User: {user_message}\nAssistant: {assistant_response}\n"
+    conversation += f"User: {message}\nAssistant:"
+
+    # Fixed values for generation parameters
+    max_tokens = 100  # Reduce max tokens if possible
+    temperature = 0.7
+    top_p = 0.9
+
+    response = generate_response(conversation, model, tokenizer, max_length=max_tokens, temperature=temperature, top_p=top_p)
     
-    return [(message, response)]
+    return response
 
-# Create the Gradio chat interface
+# Gradio Chat Interface without customizable inputs
 demo = gr.ChatInterface(
-    fn=generate_response,
-  
-    title="Chatbot",
-    description="Ask anything to the chatbot."
+    respond
 )
 
 if __name__ == "__main__":