diff --git a/compose-vLLM.yml b/compose-vLLM.yml index 7c99965..088ddc4 100644 --- a/compose-vLLM.yml +++ b/compose-vLLM.yml @@ -35,9 +35,9 @@ services: - "--dtype" - "half" - "--gpu-memory-utilization" - - "0.95" + - "0.90" - "--max-model-len" - - "16384" + - "65536" - "--max-num-seqs" - "32" - "--enable-prefix-caching"