minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

fix: add back priorty as radix cache policy (#23275)

Browse Source
This commit is contained in:
ishandhanani
2026-04-20 10:04:35 -07:00
committed by GitHub
parent 332ec5e5ee
commit b5d9a86e4c
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/server_args.py
+2 -2
View File
@@ -201,7 +201,7 @@ FP4_GEMM_RUNNER_BACKEND_CHOICES = [
"flashinfer_trtllm", "flashinfer_trtllm",
] ]
RADIX_EVICTION_POLICY_CHOICES = ["lru", "lfu", "slru"] RADIX_EVICTION_POLICY_CHOICES = ["lru", "lfu", "slru", "priority"]
RL_ON_POLICY_TARGET_CHOICES = ["fsdp"] RL_ON_POLICY_TARGET_CHOICES = ["fsdp"]
@@ -4391,7 +4391,7 @@ class ServerArgs:
type=str, type=str,
choices=RADIX_EVICTION_POLICY_CHOICES, choices=RADIX_EVICTION_POLICY_CHOICES,
default=ServerArgs.radix_eviction_policy, 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-priority requests first.",
) )
parser.add_argument( parser.add_argument(
"--enable-prefill-delayer", "--enable-prefill-delayer",