forked from docs/doc-exports
Lai, Weijian
6aa966a79a
Reviewed-by: Pruthi, Vineet <vineet.pruthi@t-systems.com> Co-authored-by: Lai, Weijian <laiweijian4@huawei.com> Co-committed-by: Lai, Weijian <laiweijian4@huawei.com>
1.7 KiB
1.7 KiB
How Do I Use Multiple Ascend Cards for Debugging in a Notebook Instance?
An Ascend multi-card training job runs in multi-process, multi-card mode. The number of cards is equal to the number of Python processes. The Ascend underlayer reads the environment variable RANK_TABLE_FILE, which has been configured in the development environment, without requiring manual configuration. For example, to run a job on eight cards, the code is as follows:
export RANK_SIZE=8
current_exec_path=$(pwd)
echo 'start training'
for((i=0;i<=$RANK_SIZE-1;i++));
do
echo 'start rank '$i
mkdir ${current_exec_path}/device$i
cd ${current_exec_path}/device$i
echo $i
export RANK_ID=$i
dev=`expr $i + 0`
echo $dev
export DEVICE_ID=$dev
python train.py > train.log 2>&1 &
done
Set the environment variable DEVICE_ID in train.py.
devid = int(os.getenv('DEVICE_ID'))
context.set_context(mode=context.GRAPH_MODE, device_target="Ascend", device_id=devid)
Parent topic: Others