diff --git a/infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml b/infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml index d9ff93a..ec44fc3 100644 --- a/infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml +++ b/infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml @@ -40,7 +40,7 @@ spec: - --gpu-memory-utilization=0.90 - --max-model-len=262144 - --enable-auto-tool-choice - - --kv-cache-dtype=bfloat16 + - --kv-cache-dtype=auto - --max-num-batched-tokens=16384 - --max-num-seqs=4 - --block-size=32