[HICache] introduce evict policy (#10190)
Signed-off-by: Xuchun Shang <xuchun.shang@linux.alibaba.com> Co-authored-by: Teng Ma <sima.mt@alibaba-inc.com>
This commit is contained in:
@@ -667,6 +667,7 @@ class Scheduler(
|
||||
else self.tp_cpu_group
|
||||
),
|
||||
page_size=self.page_size,
|
||||
eviction_policy=server_args.radix_eviction_policy,
|
||||
hicache_ratio=server_args.hicache_ratio,
|
||||
hicache_size=server_args.hicache_size,
|
||||
hicache_write_policy=server_args.hicache_write_policy,
|
||||
@@ -719,6 +720,7 @@ class Scheduler(
|
||||
tp_size=self.tp_size,
|
||||
rank=self.tp_rank,
|
||||
tp_group=self.tp_group,
|
||||
eviction_policy=server_args.radix_eviction_policy,
|
||||
)
|
||||
else:
|
||||
self.tree_cache = RadixCache(
|
||||
@@ -727,6 +729,7 @@ class Scheduler(
|
||||
page_size=self.page_size,
|
||||
disable=server_args.disable_radix_cache,
|
||||
enable_kv_cache_events=self.enable_kv_cache_events,
|
||||
eviction_policy=server_args.radix_eviction_policy,
|
||||
)
|
||||
|
||||
self.decode_mem_cache_buf_multiplier = (
|
||||
|
||||
Reference in New Issue
Block a user