Update app.py

app.py

@@ -18,9 +18,8 @@ max_seq_length = 2048
 tokenizer = AutoTokenizer.from_pretrained("ua-l/gemma-2-9b-legal-steps200-merged-16bit-uk")
 model = AutoModelForCausalLM.from_pretrained(
     "ua-l/gemma-2-9b-legal-steps200-merged-16bit-uk",
-    device='cuda',
     torch_dtype=torch.bfloat16,
-)
+).to('cuda')
 # compiled_model = torch.compile(model, mode="default")