more updates

app.py

@@ -23,7 +23,7 @@ tokenizer = AutoTokenizer.from_pretrained(mistral_path)
 tokenizer.pad_token_id = 0
 
 global model
-
+global cuda
 
 global isModelDefined
 isModelDefined = False
@@ -31,6 +31,7 @@ isModelDefined = False
 def defineModel():
     global model
     global isModelDefined
+    global cuda
     cuda = torch.cuda.is_available()
     if not isModelDefined:
         model = AutoModelForCausalLM.from_pretrained(
@@ -164,6 +165,7 @@ def generate_response(system_prompt, user_message, history, max_new_tokens, repi
     }
 
     timeout = 120.0
+    global cuda
     if cuda:
         timeout = 15.0
     _streamer = TextIteratorStreamer(tokenizer, timeout=timeout, skip_prompt=True, skip_special_tokens=False,)
@@ -488,10 +490,8 @@ with gr.Blocks(
         label="Use"
     ):
         # Header
-        if cuda:
-            gr.Markdown("# 🧠 LLM Mind Control")
-        else:
-            gr.Markdown("""# 🧠 LLM Mind Control
+
+        gr.Markdown("""# 🧠 LLM Mind Control
 
     *Warning: this space won't work well on CPU. Use the [Llama 1B version](https://huggingface.co/spaces/Abrak/Controlled_Chat_CPU) instead, or duplicate this space onto GPU hardware.""")
         gr.Markdown("""Unlike prompting, direct weight manipulation lets you fine-tune the amount of a personality