Remove unused imports in prefill.py, decode.py, and npu_graph_runner.py

- Comes from merge conflicts

Author: CatherineSue

python/sglang/srt/disaggregation/decode.py

@@ -25,7 +25,7 @@
 from collections import deque
 from dataclasses import dataclass
 from http import HTTPStatus
-from typing import TYPE_CHECKING, Dict, List, Optional, Tuple, Type, Union
+from typing import TYPE_CHECKING, List, Optional, Tuple, Type, Union
 
 import torch
 from torch.distributed import ProcessGroup
@@ -48,10 +48,7 @@
 )
 from sglang.srt.layers.dp_attention import get_attention_tp_size
 from sglang.srt.managers.schedule_batch import FINISH_ABORT, RequestStage, ScheduleBatch
-from sglang.srt.mem_cache.allocator import (
-    BaseTokenToKVPoolAllocator,
-    SWATokenToKVPoolAllocator,
-)
+from sglang.srt.mem_cache.allocator import BaseTokenToKVPoolAllocator
 from sglang.srt.mem_cache.base_prefix_cache import BasePrefixCache
 from sglang.srt.mem_cache.memory_pool import (
     HybridLinearKVPool,
@@ -61,7 +58,6 @@
     ReqToTokenPool,
     SWAKVPool,
 )
-from sglang.srt.model_executor.forward_batch_info import ForwardMode
 from sglang.srt.utils import get_int_env_var, require_mlp_sync
 from sglang.srt.utils.torch_memory_saver_adapter import TorchMemorySaverAdapter
 

python/sglang/srt/disaggregation/prefill.py

@@ -20,7 +20,6 @@
 from __future__ import annotations
 
 import logging
-import threading
 import time
 from collections import deque
 from http import HTTPStatus
@@ -54,7 +53,7 @@
     NSATokenToKVPool,
     SWAKVPool,
 )
-from sglang.srt.model_executor.forward_batch_info import ForwardMode, PPProxyTensors
+from sglang.srt.model_executor.forward_batch_info import PPProxyTensors
 from sglang.srt.utils import (
     DynamicGradMode,
     broadcast_pyobj,

python/sglang/srt/model_executor/npu_graph_runner.py

@@ -19,10 +19,9 @@
 import threading
 from typing import TYPE_CHECKING, Optional, Union
 
-import numpy as np
 import torch
 
-from sglang.srt.configs.model_config import AttentionArch, is_deepseek_nsa
+from sglang.srt.configs.model_config import is_deepseek_nsa
 from sglang.srt.model_executor.cuda_graph_runner import CudaGraphRunner
 
 logger = logging.getLogger(__name__)