Greatly thank ZhangSu's code and tutorial for the detailed instruction on NSCC.
- change
PROJECT_CODE
andTASK_NAME
in assets/g*.pbs to your own - Change the location of sif and arguments in your main.py accordingly
- Refer to
nscc_ZhangSu.md
for more details
python3 batch_qsub.py
to submit lots of jobswatch -n 10 ./watch_jobs.sh
to check the status of all the jobs./cancel_all_jobs.sh
to cancel all the jobs
batch submission of jobs with different arguments on one GPU node
import os
for dataset in ["Dataset1", "Dataset2"]:
for model in ["model1", "model2"]:
for arg in ["arg1", 'arg2']:
cmd = f"qsub -v dataset={dataset},model={model},arg={arg} asset/g1.pbs"
os.system(cmd)
print(cmd)