Update app.py

app.py

@@ -14,11 +14,7 @@ llm = Llama(
     n_ctx=2048,
     chat_format="llama-3",
     n_gpu_layers=-1,  # ensure all layers are on GPU
-    n_threads=1, # no CPU multi-threading
-    offload_kqv=True, # store kqv on GPU
-    vocab_only=False,
-    use_mmap=True,
-    use_mlock=False,
+    split_mode="LLAMA_SPLIT_MODE_NONE",
 )
 
 # Placeholder responses for when context is empty