Add example of using MLflow or Weights & Biases

[mdiazmel]

External tools can be used to manage ML/DL experiences (track hyperparameters, visualize metrics, store models, etc...).
Some of them are MLFlow, Weights & Biases and in a different context Hydra.

Some Jean Zay user's have been experiencing with these tools, having some examples (or maybe some kind of benchmarking) on how to use them in JZ should be helpful for other people.

[zaccharieramzi]

I personally have been in the last few days:

• I have setup W&B to run offline and be synced from a front node: see this script.
• I just started using Hydra with the submitit launcher. I find it very useful and elegant, and I am definitely willing to write an example combining it all together for jean-zay-doc (probably in TF/Keras but this shouldn't be too difficult to translate to PyTorch).

[mdiazmel]

Nice! PR #79 partially solves this issue. Thanks @zaccharieramzi

I'll try to add an example in the next weeks to illustrate the usage of MLFlow.