# vllm: 0.28.0 -> 0.33.0, for this configuration  (input read as: docker)

None of the options set in this configuration changed.

## Defaults that changed for options this configuration does not set (5)

option                         change
-----------------------------  ----------------------------------------
--allowed-origins              default ['*'] -> []
--cp-kv-cache-interleave-size  default 1 -> unset
--dcp-comm-backend             default 'ag_rs' -> unset
--disable-cascade-attn         default True -> unset; negations added: --enable-cascade-attn
--mamba-cache-mode             default 'none' -> unset; choices removed: all

## Suggested configuration for vllm 0.33.0

docker run \
  --gpus all \
  --ipc=host \
  -p 8000:8000 \
  -v ~/.cache/huggingface:/root/.cache/huggingface \
  -e HF_TOKEN=$HF_TOKEN \
  vllm/vllm-openai:v0.33.0 \
  --model meta-llama/Llama-3.1-8B-Instruct \
  --dtype bfloat16 \
  --max-model-len 16384 \
  --max-num-seqs 64 \
  --seed 7 \
  --trust-remote-code \
  --download-dir /root/.cache/huggingface
