|
|
|
[<img width="800px" src="https://gricad-gitlab.univ-grenoble-alpes.fr/talks/fidle/raw/master/fidle/img/00-Fidle-header-01.svg"></img>](#)
|
|
|
|
|
|
|
|
## Running Jupyter Lab on Jean-Zay
|
|
|
|
|
|
|
|
Here is a quick procedure to start Jupyter lab on a Jean-Zay node.
|
|
|
|
This procedure is intended for training given at IDRIS.
|
|
|
|
For more information, you can consult the excellent online documentation offered by IDRIS :
|
|
|
|
- [All the documentation about **Jean Zay**](http://www.idris.fr/jean-zay/)
|
|
|
|
- [Jupyter lab and Tensorboard](http://www.idris.fr/jean-zay/pre-post/jean-zay-jupyter-notebook.html)
|
|
|
|
|
|
|
|
## Special Step - Short Way, during Fidle session !
|
|
|
|
|
|
|
|
```
|
|
|
|
[jean-zay2]$ interactif
|
|
|
|
[bash]$ jupyter_start
|
|
|
|
```
|
|
|
|
|
|
|
|
Outside Fidle session, here are the step by step to get GPU resource and launch jupyter lab :
|
|
|
|
## Step 1 - Get a computer node :
|
|
|
|
|
|
|
|
Get a GPU node for 1 hour, with `salloc` command :
|
|
|
|
```
|
|
|
|
$ salloc --ntasks=1 --cpus-per-task=10 --gres=gpu:1 --hint=nomultithread --time=01:00:00 srun --pty bash
|
|
|
|
```
|
|
|
|
|
|
|
|
## Step 2 - Load the environment of your dreams :
|
|
|
|
```
|
|
|
|
$ module load tensorflow-gpu/py3/2.2.0
|
|
|
|
$ jupyter tensorboard enable --user
|
|
|
|
```
|
|
|
|
|
|
|
|
## Step 3 - Start Jupyter !
|
|
|
|
|
|
|
|
```
|
|
|
|
idrlab --notebook-dir=$WORK
|
|
|
|
```
|
|
|
|
|
|
|
|
## Step 4 - On your browser
|
|
|
|
```
|
|
|
|
go to : https://jean-zay-srv2.idris.fr
|
|
|
|
```
|
|
|
|
|
