Instructions and examples for running jobs on Kale or Ukko2¶
Example jobs¶
The following commands need to be run every time you log in to one of the clusters:
module load Python/3.7.0-intel-2018b
export SCIKIT_LEARN_DATA=$TMPDIR
cd $WRKDIR/dpEmu
source venv/bin/activate
Running text classification example¶
Create the batch file for the job:
nano text_classification.job
Then write the following content to it and save the file. Remember to put your username in place of <username>:
#!/bin/bash
#SBATCH -J dpEmu
#SBATCH --workdir=/wrk/users/<username>/dpEmu/
#SBATCH -o text_classification_results.txt
#SBATCH -c 8
#SBATCH --mem=64G
#SBATCH -t 10:00
srun python3 examples/run_text_classification_example.py all 10
srun sleep 60
Submit the batch job to be run:
sbatch text_classification.job
You can view the execution of the code as if it was executed on your home terminal with:
tail -f text_classification_results.txt
The resulting images will saved to the dpEmu/out directory.
Running object detection example¶
First remember to load the required modules and install the object detection example requirements while in the virtual enviroment, if not done already: Object detection example and notebooks requirements.
Create the batch file for the job:
nano object_detection.job
Then write the following content to it and save the file. Remember to put your username in place of <username>:
#!/bin/bash
#SBATCH -J dpEmu
#SBATCH --workdir=/wrk/users/<username>/dpEmu/
#SBATCH -o object_detection_results.txt
#SBATCH -c 4
#SBATCH --mem=32G
#SBATCH -p gpu
#SBATCH --gres=gpu:1
#SBATCH -t 10:00:00
srun python3 examples/run_object_detection_example.py
srun sleep 60
Running this example can take a lot of time. You could try to disable some of the slowest models i.e. FasterRCNN and RetinaNet. To further speed up the job on Kale, by using the latest GPUs, add the following line to the batch file:
#SBATCH --constraint=v100
Submit the batch job to be run:
sbatch object_detection.job
You can view the execution of the code as if it was executed on your home terminal with:
tail -f object_detection_results.txt
The resulting images will saved to the dpEmu/out directory.
Running object detection notebook¶
In the batch file replace:
srun python3 examples/run_object_detection_example.py
with for example:
srun jupyter nbconvert --to notebook --ExecutePreprocessor.timeout=None --inplace --execute docs/case_studies/Object_Detection_JPEG_Compression.ipynb