Prérequis

If you’re running Linux, it’s highly likely that Docker is already installed!
If not, we recommend that you install Docker Desktop.
- Check the Download Docker Desktop page info
- Do not forget the post-installation steps (for Linux users)

Option 1 - With Docker desktop, in vidéo :

Docker Desktop is a graphical interface for managing Docker environments, availaible for Linux, Windows and MacOS.
Once you’ve installed Docker Desktop, here’s how to retrieve and use the Docker Fidle image :-)

Don’t hesitate to consult documentation if you have any questions or would like to find out more!

Option 2 - Classic installation with Docker, step by step

Note :

You must have Docker installed (see prerequisites).
All commands presented below and prefixed by $ are to be launched on a terminal/unix console :smile:

1 - Get the Fidle Docker image

Get Fidle image : …it takes some time… (~5mn) and the image is quiet huge (~3 GB)
```
$ docker pull fidlehub/fidle:latest
```

To see if your image is there:

$ docker images
On standard output, you should have a line of the form :
 fidlehub/fidle    latest    [...]    3 GB 

2 - Create your fidle container

$ docker run -i  -p 8888:8888 -p 6006:6006 --name fidle  fidlehub/fidle:latest

Where:
--name fidle is used to give a name of your choice to your container
-i is used to stay in interactive mode (Keep STDIN open even if not attached)
-p Publish a container’s port(s) to the host -p Publish a container’s port to the tensorboard

3 - Open your jupyter lab

Copy the URL printed on screen beginning with http://127.0.0.1:8888/?token…
Open a Web Browser and paste the URL
Jupyterlab for fidle should be there !

4- Tensorboard

If you want to use tensoboard inside jupyter lab use this command from a terminal :

$ tensorboard --host 0.0.0.0 --logdir <path/to/your/logs>

5 - To stop your jupyter, you can

In Jupyter lab menu : “File”/”Shutdown”
On the Unix console/Terminal you launched Fidle Jupyterlab:
use Control-C to stop this server and shutdown all kernels

6 - To restart your fidle container where you leave your work

you can Just type:

$ docker start -i fidle

How to update your Docker Container ?

It is possible to retrieve a new version of the notebooks from your container.

Of course, this only applies if your container’s Python environment is compatible with the new notebooks, which is generally the case during the season.

Here’s the 4-step procedure: (need 3 minutes max)

Step 1 - Ouvrir un terminal

Once you have started your container and launched Jupyter Lab, go to “File / New / Terminal”.
A new “Terminal” tab should open.

Step 2 - Check you are at the right place !

In the Terminal tab, type the following command : ls
You should have something like that :

(...):/notebooks# ls
fidle-master-3.0.5  last
(...):/notebooks# 

Note :

If the “last” item does not appear, we advise you to retrieve a new image directly.
Dans l’exemple, votre dernière version est la 3.0.5
Important : Check that you don’t already have the latest available version.

Step 3 - Retrieve new notebooks

Still from the terminal, type the following command: fid install_notebooks --quiet

You should have something like :

(...):/notebooks# fid install_notebooks --quiet

==========================================================
fid - Your favorite Fidle admin command :-)    (v2.3.0)
==========================================================

Install Fidle notebooks in . : 

Install ressource : fidle-master
In directory      : .
Extract  :       [########################################] 100.0% of 151 files
Installed in : ./fidle-master-3.0.9
Done.

Note : If a message indicates that you already had the latest version, don’t continue.

Step 4 - Switch last version to the new one

Still from the terminal, type the following command: rm last && ln -s $(ls -1dt fidle-* | head -1) last

You should have something like :

(...):/notebooks# rm last && ln -s $(ls -1dt fidle-* | head -1) last
(...):/notebooks# 

Note :

Copy and paste this command to avoid any errors, as this could corrupt your container.
If this happens, recreate a new container and repeat the procedure…
or retrieve a more recent image directly !

Option : Access all older versions of notebooks

As an option, you can add a shortcut to all notebook versions with the following command (always from your terminal):

(...):/notebooks# ln -s /notebooks last/__Notebooks__

You should see a “Notebooks” folder in your Jupyter Lab tree :-) Have fun !