33SCRIPT_DIR=" $( cd " $( dirname " ${BASH_SOURCE[0]} " ) " && pwd) "
44
55if [[ $# -lt 1 ]]; then
6- echo " Usage: $0 <prefiller | decoder | proxy> [model]"
6+ echo " Usage: $0 <prefill | decode | proxy> [model]"
77 exit 1
88fi
99
1616fi
1717
1818
19- if [[ $1 == " prefiller " ]]; then
19+ if [[ $1 == " prefill " ]]; then
2020 # Prefiller listens on port 8100
2121 prefill_config_file=$SCRIPT_DIR /configs/lmcache-prefiller-config.yaml
2222
@@ -25,7 +25,6 @@ if [[ $1 == "prefiller" ]]; then
2525 LMCACHE_USE_EXPERIMENTAL=True \
2626 VLLM_ENABLE_V1_MULTIPROCESSING=1 \
2727 VLLM_WORKER_MULTIPROC_METHOD=spawn \
28- CUDA_VISIBLE_DEVICES=6 \
2928 vllm serve $MODEL \
3029 --port 8100 \
3130 --enforce-eager \
@@ -36,7 +35,7 @@ if [[ $1 == "prefiller" ]]; then
3635 # LMCACHE_LOG_LEVEL=DEBUG -- Set log level to DEBUG
3736 # --enforce-eager -- Enforce eager mode
3837
39- elif [[ $1 == " decoder " ]]; then
38+ elif [[ $1 == " decode " ]]; then
4039 # Decoder listens on port 8200
4140 decode_config_file=$SCRIPT_DIR /configs/lmcache-decoder-config.yaml
4241
@@ -45,7 +44,6 @@ elif [[ $1 == "decoder" ]]; then
4544 LMCACHE_USE_EXPERIMENTAL=True \
4645 VLLM_ENABLE_V1_MULTIPROCESSING=1 \
4746 VLLM_WORKER_MULTIPROC_METHOD=spawn \
48- CUDA_VISIBLE_DEVICES=7 \
4947 vllm serve $MODEL \
5048 --port 8200 \
5149 --enforce-eager \
0 commit comments