CUDA_VISIBLE_DEVICES=1 python exercise_5_HPC.py #run code on gpu #interactive node for the class nvidia-smi #check available gpus cd /dtu/datasets1/02514/ #course datasets getquota_zhome.sh #check storage space bsub -app c02516_1g.10gb < jobscript.sh #run bash script in this course, memory limited to 10 GB, not much more is needed bsub -app c02516_2g.20gb < jobscript.sh bsub -app c02516_4g.40gb < jobscript.sh #BSUB -R "select[gpu32gb]" #request 32 GB gpu only on gpuv100 bsub -q gpua100 -J "deep learning in computer vision training" -n 4 -gpu "num=1" -W 06:00 -R "rusage[mem=16GB]" -B -N -o gpu_%J.out -e gpu_%J.err6< run.sh #run command for different queue a100sh #GPU interactive node sxm2sh #GPU interactive node voltash #GPU interactive node
Jupyter setup: remote: export PATH=$PATH:~/.local/bin jupyter notebook --no-browser --port=8888 --ip=$HOSTNAME local: ssh USER@login2.hpc.dtu.dk -g -LXXXXX:n-00-00-00:XXXXX –N
Browser: http://127.0.0.1:XXXXX/?token=xyxyxyxyxyxyxyxyxyxyxyxyyxyxyxyxyxxy
02516sh source .venv_deep_learning/bin/activate export PATH=$PATH:~/.local/bin jupyter notebook --no-browser --port=8888 --ip=$HOSTNAME