Removed Cuda

models/model_loader.py

@@ -43,7 +43,7 @@ def switch_llm(model_id: str) -> str:
 def _load_llm(model_id: str):
     """Load model + its paired tokenizer. Both come from the same model_id."""
     global _llm, _tokenizer, _current_model_id
-    _tokenizer = AutoTokenizer.from_pretrained(model_id).to("cuda")
+    _tokenizer = AutoTokenizer.from_pretrained(model_id) #.to("cuda")
     _llm = AutoModelForCausalLM.from_pretrained(
         model_id,
         torch_dtype=torch.float32,