del

app.py

@@ -15,10 +15,6 @@ model, tokenizer = FastLanguageModel.from_pretrained(
     load_in_4bit=load_in_4bit,
 )
 
-# Move model to GPU if available
-device = "cuda" if torch.cuda.is_available() else "cpu"
-model = model.to(device)
-
 
 # Respond function
 def respond(
@@ -48,7 +44,7 @@ def respond(
         tokenize=True,
         add_generation_prompt=True,
         return_tensors="pt",
-    ).to(device)
+    )
 
     # Generate the response using your model
     outputs = model.generate(