Clean up server_args, triton cache manager (#8332)

python/sglang/srt/layers/moe/topk.py

@@ -15,7 +15,7 @@
 from __future__ import annotations
 
 import math
-from typing import TYPE_CHECKING, Callable, NamedTuple, Optional
+from typing import Callable, NamedTuple, Optional
 
 import torch
 import torch.nn.functional as F
@@ -39,10 +39,10 @@ from sglang.srt.utils import (
 
 _is_cuda = is_cuda()
 _is_hip = is_hip()
-_use_aiter = get_bool_env_var("SGLANG_USE_AITER") and _is_hip
-_is_cpu_amx_available = cpu_has_amx_support()
 _is_cpu = is_cpu()
+_is_cpu_amx_available = cpu_has_amx_support()
 _is_npu = is_npu()
+_use_aiter = get_bool_env_var("SGLANG_USE_AITER") and _is_hip
 
 if _is_cuda:
     from sgl_kernel import moe_fused_gate
@@ -54,7 +54,6 @@ if _use_aiter:
         from aiter import biased_grouped_topk as aiter_biased_grouped_topk
     except ImportError:
         raise ImportError("aiter is required when SGLANG_USE_AITER is set to True")
-
 if _is_npu:
     import torch_npu