diff --git a/infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml b/infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml index 7f58b97..1943577 100644 --- a/infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml +++ b/infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml @@ -39,7 +39,7 @@ spec: - --tensor-parallel-size=1 - --tokenizer-mode=hf - --gpu-memory-utilization=0.95 - - --kv-cache-dtype=fp8_e5m2 + - --kv-cache-dtype=fp8 - --max-model-len=256389 - --enable-auto-tool-choice - --enable-chunked-prefill