gahow/agentic-pd-hybrid
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add kvcache-centric profiling and admission controls

Browse Source
This commit is contained in:
Gahow Wang
2026-04-25 16:00:52 +00:00
parent 08b13d22bc
commit 13bb31a446
9 changed files with 1044 additions and 34 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
third_party/sglang/python/sglang/srt/server_args.py vendored
View File

@@ -176,7 +176,7 @@ NSA_CHOICES = [
"trtllm",
]
RADIX_EVICTION_POLICY_CHOICES = ["lru", "lfu", "slru"]
RADIX_EVICTION_POLICY_CHOICES = ["lru", "lfu", "slru", "priority"]
RL_ON_POLICY_TARGET_CHOICES = ["fsdp"]
@@ -4049,7 +4049,7 @@ class ServerArgs:
type=str,
choices=RADIX_EVICTION_POLICY_CHOICES,
default=ServerArgs.radix_eviction_policy,
help="The eviction policy of radix trees. 'lru' stands for Least Recently Used, 'lfu' stands for Least Frequently Used, and 'slru' stands for Segmented Least Recently Used.",
help="The eviction policy of radix trees. 'lru' stands for Least Recently Used, 'lfu' stands for Least Frequently Used, 'slru' stands for Segmented Least Recently Used, and 'priority' evicts lower request priority values first.",
)
parser.add_argument(
"--enable-prefill-delayer",