runai_interactive.sh

#!/bin/bash
# Running
#	bash runai_interactive.sh
# will create a job yourname-inter which 
# * has "interactive priority"
# * uses 0.5 GPU (customizable)
# * starts a jupyter server at port 8888 with default password "hello"
# * runs for 8 hours
#
# Optionally you can give the name a suffix:
#	bash runai_interactive.sh 1
# will create yourname-inter1
#
# Before starting a new interactive job, delete the previous one:
#	runai delete yourname-inter

# Customize before using:
# * CLUSTER_USER and CLUSTER_USER_ID
# * MY_WORK_DIR
# * MY_GPU_AMOUNT - fraction of GPU memory to allocate. Our GPUs usually have 32GB, so 0.25 means 8GB and 0.5 means 16GB.
# * JUPYTER_CONFIG_DIR if you want to configure jupyter (for example change password)


CLUSTER_USER=yourname # find this by running `id -un` on iccvlabsrv
CLUSTER_USER_ID=????? # find this by running `id -u` on iccvlabsrv
CLUSTER_GROUP_NAME=CVLAB-unit # find this by running `id -gn` on iccvlabsrv
CLUSTER_GROUP_ID=11166 # find this by running `id -g` on iccvlabsrv

MY_GPU_AMOUNT=0.5
MY_WORK_DIR="/cvlabdata2/home/$CLUSTER_USER"

MY_IMAGE="ic-registry.epfl.ch/cvlab/lis/lab-python-ml:cuda11"
JUPYTER_CONFIG_DIR="/cvlabdata2/home/lis/kubernetes_example/.jupyter"
MY_CMD="cd $MY_WORK_DIR && timeout --preserve-status --kill-after=1m 8h jupyter lab --ip=0.0.0.0 --no-browser"

arg_job_suffix=$1
arg_job_name="$CLUSTER_USER-inter$arg_job_suffix"
runai_project="cvlab-$CLUSTER_USER" # per-user runai projects now

echo "Job [$arg_job_name] gpu $MY_GPU_AMOUNT"

runai submit $arg_job_name \
	-i $MY_IMAGE \
	--interactive \
	--gpu $MY_GPU_AMOUNT \
	--pvc runai-$runai_project-cvlabdata1:/cvlabdata1 \
	--pvc runai-$runai_project-cvlabdata2:/cvlabdata2 \
	--pvc runai-$runai_project-cvlabsrc1:/cvlabsrc1 \
	--large-shm \
	-e CLUSTER_USER=$CLUSTER_USER \
	-e CLUSTER_USER_ID=$CLUSTER_USER_ID \
	-e CLUSTER_GROUP_NAME=$CLUSTER_GROUP_NAME \
	-e CLUSTER_GROUP_ID=$CLUSTER_GROUP_ID \
	-e TORCH_HOME="/cvlabsrc1/cvlab/pytorch_model_zoo" \
	-e JUPYTER_CONFIG_DIR="$JUPYTER_CONFIG_DIR" \
	--command -- "/opt/lab/setup_and_run_command.sh '$MY_CMD'"


# check if succeeded
if [ $? -eq 0 ]; then
	runai describe job $arg_job_name

	echo ""
	echo "Connect - terminal:"
	echo "	runai bash $arg_job_name"
	echo "Connect - jupyter:"
	echo "	kubectl port-forward $arg_job_name-0-0 8888:8888"
	echo "	open in browser: http://localhost:8888 , default password 'hello'"
fi