# non-vllm stuff
# ==============
HF_TOKEN=
HF_CACHE_DIR=


#===================================  vLLM  ==================================#

#===== MODEL
VLLM_MODEL=nvidia/NVIDIA-Nemotron-Nano-9B-v2  # e.g.

#===== UTIL
# (0.85–0.95 is g; start low first)
VLLM_GPU_UTIL=0.90