To log into Cedar, you will need a Compute Canada (CC) account. For this Hackathon we created 40 guest accounts. Ideally, if you are a researcher from a Canadian academic institution, and you are planning to use Cedar in the future, you should apply for your own account following these instructions.
If you already have a CC account, we need to add you to the GPU reservations on Cedar. Please add your username to
this Google Doc.
If you are not eligible for an account or cannot obtain one (NVIDIA mentor, non-academic researcher), please pick one
of the guest account in this Google Doc. Simply add your name to any line that you pick so that everyone knows it is
taken. Each line contains a guest username and a password. Please write this info down on a piece of paper, as we will
not keep this list public forever.
Use an SSH client to log in to Cedar:
ssh <username>@cedar.computecanada.ca
If you are on Windows, we recommend free MobaXterm (Home Edition). If you are on Linux or Mac, open a terminal and type the SSH command there.
We have two GPU reservations for this Hackathon:
- Time:
from Feb 12th 00:00:00 to Feb 21st 23:59:00 (Pacific time) - ReservationName:
warmup-wr_gpu - Account:
def-training-wa_gpu - Nodes:
cdr[2478,2482,2485-2486] (4 nodes, 32 cores/node, 4 GPUs/node) - Maximum job runtime:
3 hours
- Time: from Feb 22nd 00:00:00 to March 3rd 23:59:00 (Pacific time)
- ReservationName: hackathon-wr_gpu
- Account: def-training-wa_gpu
- Nodes: cdr[338,340-350] (12 nodes, 24 cores/node, 4 GPUs/node)
- Maximum job runtime: 3 hours
Use either salloc or sbatch:
cd ~/scratch
salloc --account=def-training-wa_gpu --reservation=hackathon-wr_gpu --gres=gpu:1 --ntasks=1 --cpus-per-task=1 --time=1:0:0 --mem-per-cpu=4g
nvidia-smi -L # show the allocated GPU card(s) in your job
You can also specify the GPU architecture with --gres=gpu:v100l:1 (NVIDIA V100 Volta) but it should be picked as default.
You can find Cedar documentation in our wiki: check out "Getting started", "Running jobs", "Available software" sections, as well as Using GPUs on Cedar.
Details to be added.