Dvc.org: tutorials: use a simpler tutorial

[ghost]

I would change the tutorial for a very simple one, (for example, the one Kaggle uses for their intro to machine learning: https://www.kaggle.com/learn/intro-to-machine-learning – easy to follow, two scripts (featurization / training), and then branching to try another model; then compare metrics of both models. (Training take seconds instead of minutes.)

[ghost]

From discord, @shcheklein :

• it takes ~1 minute to train it
• data is very small
• memory footprint is 4GB as far as I remember
• so, I do agree that we can come with a simpler scenario, def no a top priority right now

[shcheklein]

To clarify, that message above is about get started, not tutorial :)

[jorgeorpinel]

The message from Ivan on this was:

...no need to change tutorial, there is a room for multiple tutorials.

[ryokugyu]

@mroutis here is a link for tutorial created by me.
data is small
it takes 30-40 seconds to train.
https://github.com/ryokugyu/dvc_tutorial

[shcheklein]

@ryokugyu can you find some time and finish the repo with your tutorial? It's hard to understand what's going on - I mean even what problem do we exactly solve? There are some leftovers (like pyc), etc. Then we can create a blog post out of it and put a link to the documentation.

[jorgeorpinel]

Is this issue basically the same as #546 build interactive lessons with katacoda now?

[shcheklein]

no, I think this issue comes from the fact that even get started requires some substantial amount of memory to run. I think, to some extent since we already have Tutorial_s_ now users have way more options, including Katacoda. May be we can specify in the NLP tutorials that they require more memory and there other options available?

[dashohoxha]

I have noticed that the MNIST tutorial runs well even on low memory (1.5G RAM). Also it has only 3 stages (featurize, train, evaluate). It also describes experimentation by branches, using methods SVM, PCA and CNN.
Maybe this tutorial should be promoted more.

[shcheklein]

We don't have the deep tutorial anymore. Only get started. We'll rely on external tutorials for other cases. It's very hard to support and keep updated too many tutorials.