Update vllm.py - set default backend to gemlite

Author: mobicham

hqq/utils/vllm.py

@@ -523,7 +523,7 @@ class VLLM_HQQ_BACKEND:
     PYTORCH = HQQPytorchConfig
 
 
-DEFAULT_VLLM_HQQ_BACKEND = VLLM_HQQ_BACKEND.MARLIN
+DEFAULT_VLLM_HQQ_BACKEND = VLLM_HQQ_BACKEND.GEMLITE
 COMPILE_OPTIONS = {}  # {"mode":"max-autotune-no-cudagraphs"}