diff --git a/.env b/.env index efbc2b0..73e18be 100644 --- a/.env +++ b/.env @@ -6,7 +6,7 @@ SERVED_NAME=deepseek-coder TP_SIZE=4 DTYPE=float16 MAX_MODEL_LEN=65536 -GPU_MEM_UTIL=0.60 # 0.88 +GPU_MEM_UTIL=0.60 MAX_NUM_SEQS=4 VLLM_EXTRA_ARGS=--quantization gptq --disable-custom-all-reduce --enforce-eager