lower num seq

2026-05-10 02:15:27 +00:00 · 2026-05-10 02:15:27 +00:00 · a718acffa6
commit a718acffa6
parent 52f1af7e73
1 changed files with 1 additions and 1 deletions
--- a/infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml
+++ b/infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml
@ -42,7 +42,7 @@ spec:
            - --enable-auto-tool-choice 
            - --kv-cache-dtype=fp8 
            - --max-num-batched-tokens=65536 
-            - --max-num-seqs=128
+            - --max-num-seqs=64 
            - --enable-chunked-prefill
            - --trust-remote-code
            - --dtype=auto