Update app.py

app.py

@@ -21,8 +21,7 @@ model_name = "Pclanglais/Bellay"
 
 tokenizer = transformers.AutoTokenizer.from_pretrained(model_name)
 model = transformers.AutoModelForCausalLM.from_pretrained(model_name,
-    device_map="auto",
-    torch_dtype=torch.bfloat16
+    device_map="auto"
 )
 
 
@@ -72,7 +71,7 @@ class MistralChatBot:
         # Combine the user and assistant messages into a conversation
         conversation = "<|im_start|>system\n" + system_prompt + "<|im_end|>\n<|im_start|>user\n" + user_message + "<|im_end|>\n<|im_start|>assistant\n"
         # Encode the conversation using the tokenizer
-        input_ids = tokenizer.encode(conversation, return_tensors="pt", add_special_tokens=False)
+        input_ids = tokenizer.encode(conversation, return_tensors="pt", add_special_tokens=True)
         input_ids = input_ids.to(device)
         # Generate a response using the Falcon model
         response = model.generate(