Update rtx6000-vllm.yaml

2026-05-24 15:52:52 -04:00 · 2026-05-24 15:52:52 -04:00 · 2b9630d8dc
commit 2b9630d8dc
parent b536367aea
1 changed files with 2 additions and 2 deletions
--- a/infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml
+++ b/infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml
@ -38,11 +38,11 @@ spec:
            - --tensor-parallel-size=1
            - --tokenizer-mode=hf 
            - --gpu-memory-utilization=0.90
-            - --max-model-len=131072
+            - --max-model-len=262144
            - --enable-auto-tool-choice 
            - --kv-cache-dtype=bfloat16
            - --max-num-batched-tokens=16384
-            - --max-num-seqs=64
+            - --max-num-seqs=4
            - --block-size=32 
            - --language-model-only
            - --enable-chunked-prefill