[mem_cache] Replace is_chunk_cache / is_tree_cache with supports_prefix_sharing (#42362)

This commit is contained in:
Liangsheng Yin
2026-10-03 02:07:55 -07:00
committed by GitHub
parent 8293f9af54
commit 122b59152c
20 changed files with 48 additions and 48 deletions
@@ -141,7 +141,7 @@ class BaseTokenToKVPoolAllocator(abc.ABC):
from sglang.srt.mem_cache.base_prefix_cache import EvictParams
from sglang.srt.mem_cache.common import _evict_until_allocatable
if tree_cache is None or tree_cache.is_chunk_cache():
if tree_cache is None or not tree_cache.supports_prefix_sharing():
return
shortfall = num_tokens - self.available_size()
if shortfall > 0:
+1 -1
View File
@@ -199,7 +199,7 @@ class SWATokenToKVPoolAllocator(BaseTokenToKVPoolAllocator):
def evict_to_free_tokens(self, tree_cache, num_tokens: int) -> None:
from sglang.srt.mem_cache.base_prefix_cache import EvictParams
if tree_cache is None or tree_cache.is_chunk_cache():
if tree_cache is None or not tree_cache.supports_prefix_sharing():
return
full_shortfall = max(0, num_tokens - self.full_available_size())
swa_shortfall = max(0, num_tokens - self.swa_available_size())
@@ -1191,7 +1191,7 @@ class UnifiedSWATokenToKVPoolAllocator(UnifiedSWAAllocatorBase):
) -> bool | None:
from sglang.srt.mem_cache.base_prefix_cache import EvictParams
if tree_cache is None or tree_cache.is_chunk_cache():
if tree_cache is None or not tree_cache.supports_prefix_sharing():
return
required_swa = num_tokens if swa_num_tokens is None else swa_num_tokens
reclaim_plan = self.reclaim_plan(