biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[CI][BugFix] ShellCheck cleanup to remove baseline and preserve runtime behavior (#34514)

Browse Source 
Signed-off-by: junuxyz <216036880+junuxyz@users.noreply.github.com>
This commit is contained in:
junuxyz
2026-02-17 21:22:56 +09:00
committed by GitHub
parent 28bffe9466
commit c61a98f529
55 changed files with 338 additions and 464 deletions
Download Patch File Download Diff File Expand all files Collapse all files

30
examples/online_serving/disaggregated_encoder/disagg_1e1p1d_example.sh
View File

@@ -8,7 +8,7 @@ declare -a PIDS=()
###############################################################################
MODEL="${MODEL:-Qwen/Qwen2.5-VL-3B-Instruct}"
LOG_PATH="${LOG_PATH:-./logs}"
mkdir -p $LOG_PATH
mkdir -p "$LOG_PATH"
ENCODE_PORT="${ENCODE_PORT:-19534}"
PREFILL_PORT="${PREFILL_PORT:-19535}"
@@ -84,10 +84,10 @@ trap cleanup TERM
# clear previous cache
echo "remove previous ec cache folder"
rm -rf $EC_SHARED_STORAGE_PATH
rm -rf "$EC_SHARED_STORAGE_PATH"
echo "make ec cache folder"
mkdir -p $EC_SHARED_STORAGE_PATH
mkdir -p "$EC_SHARED_STORAGE_PATH"
###############################################################################
# Encoder worker
@@ -100,7 +100,7 @@ CUDA_VISIBLE_DEVICES="$GPU_E" vllm serve "$MODEL" \
--no-enable-prefix-caching \
--max-num-batched-tokens 114688 \
--max-num-seqs 128 \
--allowed-local-media-path ${GIT_ROOT}/tests/v1/ec_connector/integration \
--allowed-local-media-path "${GIT_ROOT}"/tests/v1/ec_connector/integration \
--ec-transfer-config '{
"ec_connector": "ECExampleConnector",
"ec_role": "ec_producer",
@@ -124,7 +124,7 @@ vllm serve "$MODEL" \
--enforce-eager \
--enable-request-id-headers \
--max-num-seqs 128 \
--allowed-local-media-path ${GIT_ROOT}/tests/v1/ec_connector/integration \
--allowed-local-media-path "${GIT_ROOT}"/tests/v1/ec_connector/integration \
--ec-transfer-config '{
"ec_connector": "ECExampleConnector",
"ec_role": "ec_consumer",
@@ -152,7 +152,7 @@ vllm serve "$MODEL" \
--enforce-eager \
--enable-request-id-headers \
--max-num-seqs 128 \
--allowed-local-media-path ${GIT_ROOT}/tests/v1/ec_connector/integration \
--allowed-local-media-path "${GIT_ROOT}"/tests/v1/ec_connector/integration \
--kv-transfer-config '{
"kv_connector": "NixlConnector",
"kv_role": "kv_consumer"
@@ -162,9 +162,9 @@ vllm serve "$MODEL" \
PIDS+=($!)
# Wait for workers
wait_for_server $ENCODE_PORT
wait_for_server $PREFILL_PORT
wait_for_server $DECODE_PORT
wait_for_server "$ENCODE_PORT"
wait_for_server "$PREFILL_PORT"
wait_for_server "$DECODE_PORT"
###############################################################################
# Proxy
@@ -179,7 +179,7 @@ python disagg_epd_proxy.py \
PIDS+=($!)
wait_for_server $PROXY_PORT
wait_for_server "$PROXY_PORT"
echo "All services are up!"
###############################################################################
@@ -187,14 +187,14 @@ echo "All services are up!"
###############################################################################
echo "Running benchmark (stream)..."
vllm bench serve \
--model $MODEL \
--model "$MODEL" \
--backend openai-chat \
--endpoint /v1/chat/completions \
--dataset-name hf \
--dataset-path lmarena-ai/VisionArena-Chat \
--seed 0 \
--num-prompts $NUM_PROMPTS \
--port $PROXY_PORT
--num-prompts "$NUM_PROMPTS" \
--port "$PROXY_PORT"
PIDS+=($!)
@@ -202,10 +202,10 @@ PIDS+=($!)
# Single request with local image
###############################################################################
echo "Running single request with local image (non-stream)..."
curl http://127.0.0.1:${PROXY_PORT}/v1/chat/completions \
curl http://127.0.0.1:"${PROXY_PORT}"/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "'${MODEL}'",
"model": "'"${MODEL}"'",
"messages": [
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": [

26
examples/online_serving/disaggregated_encoder/disagg_1e1pd_example.sh
View File

@@ -8,7 +8,7 @@ declare -a PIDS=()
###############################################################################
MODEL="${MODEL:-Qwen/Qwen2.5-VL-3B-Instruct}"
LOG_PATH="${LOG_PATH:-./logs}"
mkdir -p $LOG_PATH
mkdir -p "$LOG_PATH"
ENCODE_PORT="${ENCODE_PORT:-19534}"
PREFILL_DECODE_PORT="${PREFILL_DECODE_PORT:-19535}"
@@ -78,10 +78,10 @@ trap cleanup TERM
# clear previous cache
echo "remove previous ec cache folder"
rm -rf $EC_SHARED_STORAGE_PATH
rm -rf "$EC_SHARED_STORAGE_PATH"
echo "make ec cache folder"
mkdir -p $EC_SHARED_STORAGE_PATH
mkdir -p "$EC_SHARED_STORAGE_PATH"
###############################################################################
# Encoder worker
@@ -94,7 +94,7 @@ CUDA_VISIBLE_DEVICES="$GPU_E" vllm serve "$MODEL" \
--no-enable-prefix-caching \
--max-num-batched-tokens 114688 \
--max-num-seqs 128 \
--allowed-local-media-path ${GIT_ROOT}/tests/v1/ec_connector/integration \
--allowed-local-media-path "${GIT_ROOT}"/tests/v1/ec_connector/integration \
--ec-transfer-config '{
"ec_connector": "ECExampleConnector",
"ec_role": "ec_producer",
@@ -115,7 +115,7 @@ CUDA_VISIBLE_DEVICES="$GPU_PD" vllm serve "$MODEL" \
--enforce-eager \
--enable-request-id-headers \
--max-num-seqs 128 \
--allowed-local-media-path ${GIT_ROOT}/tests/v1/ec_connector/integration \
--allowed-local-media-path "${GIT_ROOT}"/tests/v1/ec_connector/integration \
--ec-transfer-config '{
"ec_connector": "ECExampleConnector",
"ec_role": "ec_consumer",
@@ -128,8 +128,8 @@ CUDA_VISIBLE_DEVICES="$GPU_PD" vllm serve "$MODEL" \
PIDS+=($!)
# Wait for workers
wait_for_server $ENCODE_PORT
wait_for_server $PREFILL_DECODE_PORT
wait_for_server "$ENCODE_PORT"
wait_for_server "$PREFILL_DECODE_PORT"
###############################################################################
# Proxy
@@ -144,7 +144,7 @@ python disagg_epd_proxy.py \
PIDS+=($!)
wait_for_server $PROXY_PORT
wait_for_server "$PROXY_PORT"
echo "All services are up!"
###############################################################################
@@ -152,14 +152,14 @@ echo "All services are up!"
###############################################################################
echo "Running benchmark (stream)..."
vllm bench serve \
--model $MODEL \
--model "$MODEL" \
--backend openai-chat \
--endpoint /v1/chat/completions \
--dataset-name hf \
--dataset-path lmarena-ai/VisionArena-Chat \
--seed 0 \
--num-prompts $NUM_PROMPTS \
--port $PROXY_PORT
--num-prompts "$NUM_PROMPTS" \
--port "$PROXY_PORT"
PIDS+=($!)
@@ -167,10 +167,10 @@ PIDS+=($!)
# Single request with local image
###############################################################################
echo "Running single request with local image (non-stream)..."
curl http://127.0.0.1:${PROXY_PORT}/v1/chat/completions \
curl http://127.0.0.1:"${PROXY_PORT}"/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "'${MODEL}'",
"model": "'"${MODEL}"'",
"messages": [
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": [

4
examples/online_serving/disaggregated_prefill.sh
View File

@@ -54,7 +54,7 @@ wait_for_server() {
# You can also adjust --kv-ip and --kv-port for distributed inference.
# prefilling instance, which is the KV producer
CUDA_VISIBLE_DEVICES=0 vllm serve $MODEL_NAME \
CUDA_VISIBLE_DEVICES=0 vllm serve "$MODEL_NAME" \
--host 0.0.0.0 \
--port 8100 \
--max-model-len 100 \
@@ -64,7 +64,7 @@ CUDA_VISIBLE_DEVICES=0 vllm serve $MODEL_NAME \
'{"kv_connector":"P2pNcclConnector","kv_role":"kv_producer","kv_rank":0,"kv_parallel_size":2,"kv_buffer_size":"1e9","kv_port":"14579","kv_connector_extra_config":{"proxy_ip":"'"$VLLM_HOST_IP"'","proxy_port":"30001","http_ip":"'"$VLLM_HOST_IP"'","http_port":"8100","send_type":"PUT_ASYNC"}}' &
# decoding instance, which is the KV consumer
CUDA_VISIBLE_DEVICES=1 vllm serve $MODEL_NAME \
CUDA_VISIBLE_DEVICES=1 vllm serve "$MODEL_NAME" \
--host 0.0.0.0 \
--port 8200 \
--max-model-len 100 \

2
examples/online_serving/disaggregated_serving/kv_events.sh
View File

@@ -34,7 +34,7 @@ wait_for_server() {
done" && return 0 || return 1
}
vllm serve $MODEL_NAME \
vllm serve "$MODEL_NAME" \
--port 8100 \
--max-model-len 100 \
--enforce-eager \

23
examples/online_serving/disaggregated_serving/mooncake_connector/run_mooncake_connector.sh
View File

@@ -143,7 +143,7 @@ main() {
IFS=',' read -ra BOOTSTRAP_PORT_ARRAY <<< "$BOOTSTRAP_PORTS"
IFS=',' read -ra DECODE_PORT_ARRAY <<< "$DECODE_PORTS"
proxy_param=""
proxy_args=()
# =============================================================================
# Launch Prefill Servers (X Producers)
@@ -156,12 +156,12 @@ main() {
local bootstrap_port=${BOOTSTRAP_PORT_ARRAY[$i]}
echo " Prefill server $((i+1)): GPU $gpu_id, Port $port, Bootstrap Port $bootstrap_port"
VLLM_MOONCAKE_BOOTSTRAP_PORT=$bootstrap_port CUDA_VISIBLE_DEVICES=$gpu_id vllm serve $MODEL \
--port $port \
VLLM_MOONCAKE_BOOTSTRAP_PORT=$bootstrap_port CUDA_VISIBLE_DEVICES=$gpu_id vllm serve "$MODEL" \
--port "$port" \
--kv-transfer-config \
"{\"kv_connector\":\"MooncakeConnector\",\"kv_role\":\"kv_producer\"}" > prefill$((i+1)).log 2>&1 &
PIDS+=($!)
proxy_param="${proxy_param} --prefill http://0.0.0.0:${port} $bootstrap_port"
proxy_args+=(--prefill "http://0.0.0.0:${port}" "$bootstrap_port")
done
# =============================================================================
@@ -174,12 +174,12 @@ main() {
local port=${DECODE_PORT_ARRAY[$i]}
echo " Decode server $((i+1)): GPU $gpu_id, Port $port"
CUDA_VISIBLE_DEVICES=$gpu_id vllm serve $MODEL \
--port $port \
CUDA_VISIBLE_DEVICES=$gpu_id vllm serve "$MODEL" \
--port "$port" \
--kv-transfer-config \
"{\"kv_connector\":\"MooncakeConnector\",\"kv_role\":\"kv_consumer\"}" > decode$((i+1)).log 2>&1 &
PIDS+=($!)
proxy_param="${proxy_param} --decode http://0.0.0.0:${port}"
proxy_args+=(--decode "http://0.0.0.0:${port}")
done
# =============================================================================
@@ -187,7 +187,7 @@ main() {
# =============================================================================
echo ""
echo "Starting proxy server on port $PROXY_PORT..."
python3 mooncake_connector_proxy.py $proxy_param --port $PROXY_PORT > proxy.log 2>&1 &
python3 mooncake_connector_proxy.py "${proxy_args[@]}" --port "$PROXY_PORT" > proxy.log 2>&1 &
PIDS+=($!)
# =============================================================================
@@ -196,9 +196,10 @@ main() {
echo ""
echo "Waiting for all servers to start..."
for port in "${PREFILL_PORT_ARRAY[@]}" "${DECODE_PORT_ARRAY[@]}"; do
if ! wait_for_server $port; then
if ! wait_for_server "$port"; then
echo "Failed to start server on port $port"
cleanup
# shellcheck disable=SC2317
exit 1
fi
done
@@ -209,8 +210,8 @@ main() {
# =============================================================================
# Run Benchmark
# =============================================================================
vllm bench serve --port $PROXY_PORT --seed $(date +%s) \
--backend vllm --model $MODEL \
vllm bench serve --port "$PROXY_PORT" --seed "$(date +%s)" \
--backend vllm --model "$MODEL" \
--dataset-name random --random-input-len 7500 --random-output-len 200 \
--num-prompts 200 --burstiness 100 --request-rate 2 | tee benchmark.log

15
examples/online_serving/disaggregated_serving_p2p_nccl_xpyd/disagg_example_p2p_nccl_xpyd.sh
View File

@@ -166,10 +166,10 @@ main() {
local kv_port=$((21001 + i))
echo " Prefill server $((i+1)): GPU $gpu_id, Port $port, KV Port $kv_port"
CUDA_VISIBLE_DEVICES=$gpu_id vllm serve $MODEL \
CUDA_VISIBLE_DEVICES=$gpu_id vllm serve "$MODEL" \
--enforce-eager \
--host 0.0.0.0 \
--port $port \
--port "$port" \
--tensor-parallel-size 1 \
--seed 1024 \
--dtype float16 \
@@ -194,10 +194,10 @@ main() {
local kv_port=$((22001 + i))
echo " Decode server $((i+1)): GPU $gpu_id, Port $port, KV Port $kv_port"
CUDA_VISIBLE_DEVICES=$gpu_id vllm serve $MODEL \
CUDA_VISIBLE_DEVICES=$gpu_id vllm serve "$MODEL" \
--enforce-eager \
--host 0.0.0.0 \
--port $port \
--port "$port" \
--tensor-parallel-size 1 \
--seed 1024 \
--dtype float16 \
@@ -217,9 +217,10 @@ main() {
echo ""
echo "Waiting for all servers to start..."
for port in "${PREFILL_PORT_ARRAY[@]}" "${DECODE_PORT_ARRAY[@]}"; do
if ! wait_for_server $port; then
if ! wait_for_server "$port"; then
echo "Failed to start server on port $port"
cleanup
# shellcheck disable=SC2317
exit 1
fi
done
@@ -231,8 +232,8 @@ main() {
# Run Benchmark
# =============================================================================
cd ../../../benchmarks/
vllm bench serve --port 10001 --seed $(date +%s) \
--model $MODEL \
vllm bench serve --port 10001 --seed "$(date +%s)" \
--model "$MODEL" \
--dataset-name random --random-input-len 7500 --random-output-len 200 \
--num-prompts 200 --burstiness 100 --request-rate 2 | tee benchmark.log

10
examples/online_serving/elastic_ep/bench.sh
View File

@@ -50,8 +50,8 @@ while [[ $# -gt 0 ]]; do
done
vllm bench serve \
--model $MODEL_NAME \
--host $HOST \
--port $PORT \
--num-prompts $NUM_PROMPTS \
--request-rate $REQUEST_RATE
--model "$MODEL_NAME" \
--host "$HOST" \
--port "$PORT" \
--num-prompts "$NUM_PROMPTS" \
--request-rate "$REQUEST_RATE"

12
examples/online_serving/elastic_ep/serve_deepseek_v2.sh
View File

@@ -57,15 +57,15 @@ echo "Starting vLLM server for $MODEL_NAME with data parallel size: $DATA_PARALL
export RAY_DEDUP_LOGS=0
export VLLM_USE_DEEP_GEMM=1
vllm serve $MODEL_NAME \
--data-parallel-size $DATA_PARALLEL_SIZE \
--data-parallel-size-local $DATA_PARALLEL_SIZE \
vllm serve "$MODEL_NAME" \
--data-parallel-size "$DATA_PARALLEL_SIZE" \
--data-parallel-size-local "$DATA_PARALLEL_SIZE" \
--data-parallel-backend ray \
--enforce-eager \
--enable-expert-parallel \
--enable-eplb \
--all2all-backend pplx \
--num-redundant-experts $REDUNDANT_EXPERTS \
--num-redundant-experts "$REDUNDANT_EXPERTS" \
--trust-remote-code \
--host $HOST \
--port $PORT
--host "$HOST" \
--port "$PORT"

9
examples/online_serving/multi-node-serving.sh
View File

@@ -57,8 +57,7 @@ case "$subcommand" in
# Retry until the worker node connects to the head node or the timeout expires.
for (( i=0; i < $ray_init_timeout; i+=5 )); do
ray start --address=$ray_address:$ray_port --block "${start_params[@]}"
if [ $? -eq 0 ]; then
if ray start --address="$ray_address":"$ray_port" --block "${start_params[@]}"; then
echo "Worker: Ray runtime started with head address $ray_address:$ray_port"
exit 0
fi
@@ -95,12 +94,12 @@ case "$subcommand" in
fi
# Start the Ray head node.
ray start --head --port=$ray_port "${start_params[@]}"
ray start --head --port="$ray_port" "${start_params[@]}"
# Poll Ray until every worker node is active.
for (( i=0; i < $ray_init_timeout; i+=5 )); do
active_nodes=`python3 -c 'import ray; ray.init(); print(sum(node["Alive"] for node in ray.nodes()))'`
if [ $active_nodes -eq $ray_cluster_size ]; then
active_nodes=$(python3 -c 'import ray; ray.init(); print(sum(node["Alive"] for node in ray.nodes()))')
if [ "$active_nodes" -eq "$ray_cluster_size" ]; then
echo "All ray workers are active and the ray cluster is initialized successfully."
exit 0
fi

16
examples/others/lmcache/disagg_prefill_lmcache_v1/disagg_example_nixl.sh
View File

@@ -22,11 +22,10 @@ check_hf_token() {
check_num_gpus() {
# can you check if the number of GPUs are >=2 via nvidia-smi/rocm-smi?
which rocm-smi > /dev/null 2>&1
if [ $? -ne 0 ]; then
if ! which rocm-smi > /dev/null 2>&1; then
num_gpus=$(nvidia-smi --query-gpu=name --format=csv,noheader | wc -l)
else
num_gpus=$(rocm-smi --showid | grep Instinct | wc -l)
num_gpus=$(rocm-smi --showid | grep -c Instinct)
fi
if [ "$num_gpus" -lt 2 ]; then
@@ -39,8 +38,7 @@ check_num_gpus() {
ensure_python_library_installed() {
echo "Checking if $1 is installed..."
python3 -c "import $1" > /dev/null 2>&1
if [ $? -ne 0 ]; then
if ! python3 -c "import $1" > /dev/null 2>&1; then
if [ "$1" == "nixl" ]; then
echo "$1 is not installed. Please refer to https://github.com/ai-dynamo/nixl for installation."
else
@@ -102,12 +100,12 @@ main() {
bash disagg_vllm_launcher.sh prefiller \
> >(tee prefiller.log) 2>&1 &
prefiller_pid=$!
PIDS+=($prefiller_pid)
PIDS+=("$prefiller_pid")
bash disagg_vllm_launcher.sh decoder \
> >(tee decoder.log) 2>&1 &
decoder_pid=$!
PIDS+=($decoder_pid)
PIDS+=("$decoder_pid")
python3 disagg_proxy_server.py \
--host localhost \
@@ -118,7 +116,7 @@ main() {
--decoder-port 8200 \
> >(tee proxy.log) 2>&1 &
proxy_pid=$!
PIDS+=($proxy_pid)
PIDS+=("$proxy_pid")
wait_for_server 8100
wait_for_server 8200
@@ -128,7 +126,7 @@ main() {
# begin benchmark
cd ../../../../benchmarks/
vllm bench serve --port 9000 --seed $(date +%s) \
vllm bench serve --port 9000 --seed "$(date +%s)" \
--model meta-llama/Llama-3.1-8B-Instruct \
--dataset-name random --random-input-len 7500 --random-output-len 200 \
--num-prompts 200 --burstiness 100 --request-rate 3.6 | tee benchmark.log

4
examples/others/lmcache/disagg_prefill_lmcache_v1/disagg_vllm_launcher.sh
View File

@@ -34,7 +34,7 @@ if [[ $1 == "prefiller" ]]; then
VLLM_ENABLE_V1_MULTIPROCESSING=1 \
VLLM_WORKER_MULTIPROC_METHOD=spawn \
CUDA_VISIBLE_DEVICES=0 \
vllm serve $MODEL \
vllm serve "$MODEL" \
--port 8100 \
--enforce-eager \
--kv-transfer-config \
@@ -51,7 +51,7 @@ elif [[ $1 == "decoder" ]]; then
VLLM_ENABLE_V1_MULTIPROCESSING=1 \
VLLM_WORKER_MULTIPROC_METHOD=spawn \
CUDA_VISIBLE_DEVICES=1 \
vllm serve $MODEL \
vllm serve "$MODEL" \
--port 8200 \
--enforce-eager \
--kv-transfer-config \

2
examples/pooling/embed/openai_embedding_long_text/service.sh
View File

@@ -103,7 +103,7 @@ vllm serve "$MODEL_NAME" \
--tensor-parallel-size "$GPU_COUNT" \
--enforce-eager \
--pooler-config "$POOLER_CONFIG" \
--served-model-name ${MODEL_CODE} \
--served-model-name "${MODEL_CODE}" \
--api-key "$API_KEY" \
--trust-remote-code \
--port "$PORT" \