Acces à la demande à la partition gpu

Le cluster IFB Core a desormais 3 noeuds GPU avec chacun 2x Processeur graphique NVIDIA Ampere A100:

gpu-node-01
gpu-node-02
gpu-node-03

Les accès aux gpus se font à la demande.

Pour utiliser ces noeuds vous devez donc faire une demande via ce forum afin que votre login/compte soit ajouté à la partition gpu.

Hello @nc-support ,

I would like to access the GPU for my project named "Free energy complutation". Could you please add me to the gpu partition?
Thank you in advance and have a nice day ahead!

Hello Hrishikesh.

It's done you now have access to the gpu partition

Thank you
Nicole

2 « J'aime »

Je voudrais aussi demander la partition GPU. Mon nom d'utilisateur est simeons. Mon nom d'utilisateur est simeons

Bonjour Simon,
Pour lequel de vos projets avez vous besoin de GPU ?
Nous devons associer votre login et projet à la partition gpu.

Merci Nicole

1 « J'aime »

Bonjour Nicole,

Je voudrais obtenir la partition gpu pour le projet parp.

Merci beaucoup.

Bonjour Simon, voila c'est fait.
Merci Nicole

1 « J'aime »

Bonjour @nc-support

Je voudrais utiliser GPU pour projet 'aphanoclust' utilisateur claurent

Merci en avance

Bonjour @Andrei_Kiselev
Votre compte et celui de claurent ont été ajoutés.

Merci
Nicole

Hello @nc-support ,

Is it possible to submit a job for more than 24 hrs on a GPU node?

Thanks in advance!

Hello @hrishidhondge ,

Yes, it's possible (job <= 60 days).
GPU jobs have to be launch on gpu partition (described here:: Slurm at IFB - IFB Core Cluster Documentation).

Have a good day

1 « J'aime »

Bonjour,

@nc-support Pourriez-vous associer le projet "dime_intpred" à la partition GPU, svp?

Merci,

Bonjour @azanzoni,
C'est fait. Bon calcul !

1 « J'aime »

I tried with the given partition, but still, I'm getting PartitionTimeLimit and the job is pending from that point.
Although on the given documentation it is mentioned as 60-days time limit, for GPU partition the time limit is just one day. Is it possible to increase this time limit?

Hello,

You're right. There is a time limitation on the partition (1 day...).

The GPU node is a rare ressource so it's seems to me difficult to rise this limit. We discuss about it.
For our information, how many time and ressource (number of GPU) do you need ? What kind of processing/software you use ?

Hello,

I would like to have access to the GPU partition with my project "Analysis Coffea Genomes using Deep Neural Networks". Could I be included?

Thank you!!

Have a nice day

Hello @sorozcoarias ,

Which software or kind of process do you plan to use ?

We are planning to use a software develped by us, which it's based on Keras and Tensorflow (https://github.com/simonorozcoarias/Inpactor2). This is why we need to access the GPU, because the software uses neural networks.

Hello,
@sorozcoarias you have now access to the gpu partition.
Don't forget to specify this partition and your account (--partition=gpu --account=coffea_genomes).
Have a nice day

Hello @hrishidhondge,
Sorry for the delay (we had to fix some stuff before), the gpu partition time limit is now 3 days (thanks to @haessigj).
Please tell us if it's not sufficient.

1 « J'aime »