mindspore-Ecosystem
/
mindspore
21
0
Fork 27
Code Issues Pull Requests 6 Projects Releases Wiki Activity

Optimize cpu allreduce test case

This commit is contained in:
ZPaC 2022-09-27 09:40:16 +08:00
parent 57123f816c
commit 1922113083
2 changed files with 8 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
tests/st/cpu_data_parallel/test_all_reduce/build_allreduce_net_cluster.sh
View File

@ -17,10 +17,11 @@
export MS_WORKER_NUM=8 export MS_WORKER_NUM=8
export MS_SCHED_HOST=127.0.0.1 export MS_SCHED_HOST=127.0.0.1
export MS_SCHED_PORT=$2 export MS_SCHED_PORT=$2
export GLOG_v=1
# Launch 1 scheduler. # Launch 1 scheduler.
export MS_ROLE=MS_SCHED export MS_ROLE=MS_SCHED
python3 $1 >scheduler.txt 2>&1 & python3 $1 >scheduler.log 2>&1 &
sched_pid=${!} sched_pid=${!}
echo "scheduler start success!" echo "scheduler start success!"
@ -29,7 +30,7 @@ export MS_ROLE=MS_WORKER
process_pid=() process_pid=()
for((i=0;i<8;i++)); for((i=0;i<8;i++));
do do
python3 $1 >worker_$i.txt 2>&1 & python3 $1 >worker_$i.log 2>&1 &
echo "worker ${i} start success with pid ${!}" echo "worker ${i} start success with pid ${!}"
process_pid[${i}]=${!} process_pid[${i}]=${!}
done done

5
tests/st/cpu_data_parallel/test_all_reduce/test_allreduce.py
View File

@ -33,6 +33,11 @@ def test_allreduce():
if sys.platform != 'linux': if sys.platform != 'linux':
return return
return_code = os.system("bash build_allreduce_net_cluster.sh run_allreduce.py 8119") return_code = os.system("bash build_allreduce_net_cluster.sh run_allreduce.py 8119")
if return_code != 0:
os.system(f"echo '\n**************** Worker Log ****************'")
os.system(f"grep -E 'ERROR|Error|error' -C 15 ./worker*.log")
os.system(f"echo '\n**************** Scheduler Log ****************'")
os.system(f"grep -E 'ERROR|Error|error' -C 15 ./scheduler.log")
assert return_code == 0 assert return_code == 0