update arch_name and arch_cu location

vllm/model_executor/layers/fused_moe/fused_moe.py

@@ -49,6 +49,8 @@ logger = init_logger(__name__)
 if envs.VLLM_USE_GLOBAL_CACHE13:
     moe_cache_singleton = None
     
+arch_name = torch.cuda.get_device_properties("cuda").gcnArchName.split(':')[0]
+arch_cu = torch.cuda.get_device_properties(torch.cuda.current_device()).multi_processor_count
 
 @torch.compile
 def moe_sum_reduce_torch_compile(x, out, routed_scaling_factor):
@@ -996,9 +998,7 @@ def get_config_file_name(E: int,
                          N: int,
                          dtype: Optional[str],
                          block_shape: Optional[List[int]] = None, use_nn_moe: Optional[bool] = False) -> str:
-    device_name = current_platform.get_device_name().replace(" ", "_")
-    arch_name = torch.cuda.get_device_properties("cuda").gcnArchName.split(':')[0]
-    arch_cu = torch.cuda.get_device_properties(torch.cuda.current_device()).multi_processor_count
+    # device_name = current_platform.get_device_name().replace(" ", "_")
     dtype_selector = "" if not dtype else f",dtype={dtype}"
     block_shape_selector = ("" if not block_shape or not all(block_shape) else
                             f",block_shape={block_shape}").replace(" ", "")