[mem_cache][5/N] refactor: extract host KV cache base layer into pool_host package (#27273)

test/registered/jit/test_hicache.py

@@ -6,9 +6,11 @@ import torch
 from sglang.jit_kernel.hicache import can_use_write_back_jit_kernel
 from sglang.srt.mem_cache.memory_pool import MHATokenToKVPool, MLATokenToKVPool
 from sglang.srt.mem_cache.memory_pool_host import (
-    ALLOC_MEMORY_FUNCS,
     MHATokenToKVPoolHost,
     MLATokenToKVPoolHost,
+)
+from sglang.srt.mem_cache.pool_host.common import (
+    ALLOC_MEMORY_FUNCS,
     alloc_with_pin_memory,
 )
 from sglang.srt.utils import is_cuda, is_hip, is_npu, is_xpu

test/registered/unit/managers/test_hisparse_unit.py

@@ -85,7 +85,7 @@ class TestHiSparseUnit(unittest.TestCase):
             torch.distributed.init_process_group(backend="gloo", rank=0, world_size=1)
         cls.tp_group = torch.distributed.group.WORLD
 
-        from sglang.srt.mem_cache.memory_pool_host import (
+        from sglang.srt.mem_cache.pool_host.common import (
             ALLOC_MEMORY_FUNCS,
             alloc_with_pin_memory,
         )
@@ -154,7 +154,7 @@ class TestHiSparseUnit(unittest.TestCase):
 
     @classmethod
     def tearDownClass(cls):
-        from sglang.srt.mem_cache.memory_pool_host import ALLOC_MEMORY_FUNCS
+        from sglang.srt.mem_cache.pool_host.common import ALLOC_MEMORY_FUNCS
 
         ALLOC_MEMORY_FUNCS["cuda"] = cls._original_alloc
         if torch.distributed.is_initialized():

test/registered/unit/mem_cache/test_dsa_pool_host_unit.py

@@ -5,9 +5,11 @@ import torch
 
 from sglang.srt.mem_cache.memory_pool import DSATokenToKVPool
 from sglang.srt.mem_cache.memory_pool_host import (
-    ALLOC_MEMORY_FUNCS,
     DSAIndexerPoolHost,
     MLATokenToKVPoolHost,
+)
+from sglang.srt.mem_cache.pool_host.common import (
+    ALLOC_MEMORY_FUNCS,
     alloc_with_pin_memory,
 )
 from sglang.srt.utils import is_cuda, is_hip, is_npu, is_xpu