Welcome to the FAQ page for Berzelius!
On this page, you’ll find answers to commonly asked questions about our HPC environment. These FAQs are here to provide you with troubleshooting tips and best practices to optimize your experience on our cluster.
A: This message indicates that your storage quota is almost full. You can check your project quota by nscquota
.
[xuan@berzelius0 xuan]$ nscquota
Data Files
Directory Used Quota Hard Limit Used Quota Hard Limit
/home/xuan 8.0 GiB 20.0 GiB 30.0 GiB 60429 1000000 1500000
/proj/your-berzelius-project 245.1 MiB 2.0 TiB 2.9 TiB 712 2000000 3000000
You can use ncdu
to check which folders are taking the most space in your home directory or project directory.
ncdu /home/<your username>
ncdu /proj/<your project>/users/<your username>
/home/<your username>
has a quota of only 20 GB. You’ll need to move the large directory from your home directory to your project directory, then create a symlink in your home directory pointing to its original location.For example, to move the .conda
directory to your project directory:
mv ~/.conda /proj/<your project>/users/<your username>/
ln -s /proj/<your project>/users/<your username>/.conda ~/.conda
/proj/common-datasets
. If they are, you can remove your own copy to save storage quota.
This message indicates that your file quota (the limit on the number of files) is almost full. You can check your project quota by running nscquota
.
To list the number of files in each directory:
find . -type f | cut -d '/' -f 2 | sort | uniq -c | sort -nr
Please check if the datasets you need are already hosted on Berzelius at /proj/common-datasets
. If they are, you can remove your own copy to save file quota.
Alternatively, you can reduce the number of files by using archive formats like .tar
or databases such as LMDB to package your datasets.
tar
archives, refer to this example which demonstrates how to pack ImageNet datasets into tar
files and efficiently transfer them to node local storage.
A: You can extend the time limit of a running job by using the command nsc-boost-timelimit
on the login node. Please read the guide for NSC boost-tools.
A: Please make sure that you have installed the GPU version of PyTorch/TensorFlow.
To check the installation info:
# If you installed the package via pip
pip show pytorch
# If you installed the package via conda
conda list pytorch
A: Please check out Berzelius Conda/Mamba Guide.
A: Please check out Mamba.
A: The recommended way to acknowledge access to Berzelius is:
“The computations/data handling/[SIMILAR] were/was enabled by the Berzelius resource provided by the Knut and Alice Wallenberg Foundation at the National Supercomputer Centre.”
If applicable, also add an acknowledgement for application support: “[NAME] at the National Supercomputer Centre is acknowledged for assistance concerning technical and implementational aspects [OR SIMILAR] in making the code run [OR SIMILAR] on the Berzelius resource.”
or
“We thank [NAME] at NSC for [his/her] assistance with [describe tasks such as porting, optimization, etc.] on the Berzelius, which was made possible through application support provided by National Supercomputer Centre at Linköping University.”
A: Berzelius is not designed to handle sensitive data. Bianca is currently the only NAISS resource that can be used for sensitive data storage.
Guides, documentation and FAQ.
Applying for projects and login accounts.