Updated for CPU evalgguf

app.py

@@ -1,14 +1,15 @@
 import streamlit as st
-from transformers import AutoTokenizer, AutoModelForCausalLM  
+from ggulf import GGUFModel, GGUFTokenizer
 import torch
 
-# Use a medium sized model 
-model_name = "microsoft/DialoGPT-medium"  
+model_name = "TheBloke/TinyLlama-1.1B-Chat-v0.3-GGUF"  
 
-model = AutoModelForCausalLM.from_pretrained(model_name)
-model.eval() # Make sure model is in evaluation mode
+# Load model and tokenizer
+model = GGUFModel.from_pretrained(model_name)
+tokenizer = GGUFTokenizer.from_pretrained(model_name)
 
-tokenizer = AutoTokenizer.from_pretrained(model_name)
+# Set model to eval mode
+model.eval() 
 
 st.title("Buddy Christ Chatbot")
 
@@ -16,14 +17,19 @@ user_input = st.text_input("You:", "")
 
 if user_input:
 
+  # Encode input
   inputs = tokenizer.encode(user_input, return_tensors="pt")
-  
+
+  # Generate response using GGUF
   response = model.generate(inputs, 
-                            max_length=1000, 
-                            temperature=0.8, 
-                            top_k=50,
-                            pad_token_id=tokenizer.eos_token_id)
-                            
-  response_text = tokenizer.decode(response[0], skip_special_tokens=True)
+                            max_length=1000,
+                            temperature=1.0, 
+                            top_k=10,
+                            pad_token_id=tokenizer.eos_token_id,
+                            gguf_mode=True)
+  
+  # Print and display full response
+  print(response_text)
+  response_text = tokenizer.decode(response[0])
   
   st.write("Buddy Christ:", response_text)