apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-06 09:07:10 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

update code comments

Browse Source
This commit is contained in:
kevincheng2
2024-09-02 13:48:47 +08:00
parent 8500f5dfec
commit 4bc8dc38b0
23 changed files with 750 additions and 577 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
llm/server/scripts/start_server.sh
View File

@@ -14,8 +14,8 @@ export FLAGS_gemm_use_half_precision_compute_type=0
export NVIDIA_TF32_OVERRIDE=0
# Model hyperparameters
export MP_NUM=${MP_NUM:-"1"} # GPU num
export CUDA_VISIBLE_DEVICES=${CUDA_VISIBLE_DEVICES:-"0"} # GPU
export MP_NUM=${MP_NUM:-"1"} # Number of GPUs
export CUDA_VISIBLE_DEVICES=${CUDA_VISIBLE_DEVICES:-"0"} # GPU ids
export MAX_SEQ_LEN=${MAX_SEQ_LEN:-"8192"}
export MAX_DEC_LEN=${MAX_DEC_LEN:-"2048"}
export BATCH_SIZE=${BATCH_SIZE:-"20"}
@@ -44,7 +44,6 @@ mkdir -p log
rm -rf console.log log/*
rm -rf /dev/shm/*
# 启动服务
echo "start serving ..."
tritonserver --exit-timeout-secs 100 --cuda-memory-pool-byte-size 0:0 --cuda-memory-pool-byte-size 1:0 \
@@ -55,4 +54,5 @@ tritonserver --exit-timeout-secs 100 --cuda-memory-pool-byte-size 0:0 --cuda-mem
--grpc-port=${GRPC_PORT} \
--metrics-port=${METRICS_PORT} \
--log-file log/server.log --log-info true > log/console.log 2>&1 &
echo "模型服务的启动日志，请查看" ${PWD}"/log/server.log 和 "${PWD}"/log/workerlog.0 "
echo "The logs for the model service, please check" ${PWD}"/log/server.log and "${PWD}"/log/workerlog.0"

17
llm/server/scripts/stop_server.sh
View File

@@ -3,7 +3,7 @@
pids=($(ps aux | grep -E 'tritonserver' | grep -v grep | awk '{print $2}'))
if [ ${#pids[@]} -eq 0 ]; then
echo "未找到 tritonserver 相关进程"
echo "Can not find tritonserver."
timeout=1
else
timeout=300
@@ -11,7 +11,7 @@ fi
# kill processor
for pid in "${pids[@]}"; do
echo "正在中断进程 $pid"
echo "killing $pid"
kill -2 "$pid"
done
@@ -29,9 +29,8 @@ while : ; do
elapsed_time=$((current_time - start_time))
if [ $elapsed_time -ge $timeout ]; then
echo "tritonserver进程超时未退出"
echo "强制杀死所有有关进程"
pids=$(ps auxww | grep -E "tritonserver|triton_python_backend_stub|new_infer.py|infer|multiprocessing.resource_tracker|paddle.distributed.launch|task_queue_manager|app.py|memory_log.py|spawn_main" | grep -v grep | grep -v start_both | awk '{print $2}');
echo "forcibly kill all process ..."
pids=$(ps auxww | grep -E "tritonserver|triton_python_backend_stub|infer|multiprocessing.resource_tracker|paddle.distributed.launch|task_queue_manager|app.py|spawn_main" | grep -v grep | grep -v start_both | awk '{print $2}');
echo $pids;
for pid in ${pids[@]}; do
kill -9 ${pid}
@@ -39,14 +38,14 @@ while : ; do
break
fi
pids=$(ps auxww | grep -E "tritonserver|triton_python_backend_stub|new_infer.py|multiprocessing.resource_tracker|paddle.distributed.launch|app.py|memory_log.py|spawn_main" | grep -v grep | awk '{print $2}');
pids=$(ps auxww | grep -E "tritonserver|triton_python_backend_stub|multiprocessing.resource_tracker|paddle.distributed.launch|app.py|spawn_main" | grep -v grep | awk '{print $2}');
array=($(echo "$pids" | tr ' ' '\n'))
if [ ${#array[*]} -ne 0 ]; then
echo "进程还没有清理干净, 等待清理完毕"
echo "cleaning process, please wait ..."
sleep 1
else
echo "进程已经清理干净"
echo "clean finished."
break
fi
done
@@ -65,5 +64,5 @@ for in_pid in ${health_checker_pids[@]}; do
done
echo 'end kill health checker'
echo "所有进程已终止"
echo "all process terminated."
exit 0