search

TuringLang / docs

Documentation and tutorials for the Turing language
https://turinglang.org/docs/tutorials/docs-00-getting-started/
MIT License
225 stars 97 forks source link

Add tutorials from the "Introduction to Bayesian Data Analysis" workshop from Juliacon 2021 #219

Open torfjelde opened 2 years ago

torfjelde commented 2 years ago

I implemented some tutorials for the above mentioned workshop in Turing.jl (https://github.com/kskyten/BayesWorkshop2021/), and I'm thinking combining all of them into one longer tutorial (including the 5th notebook) would be a great addition to TuringTutorials:)

This issue is just a reminder (unless someone beats me to it:) ).

torfjelde commented 2 years ago

Just making sure: okay with you if we do this @kskyten? We'll of course link to the original work, etc. in the tutorial:)

kskyten commented 2 years ago

I would like to convert the notebooks to Pluto now that it should support macros and use native plotting (see https://github.com/kskyten/BayesPlot.jl). Feel free to use the notebooks. Let's collaborate in a single place though.

torfjelde commented 2 years ago

Though Pluto sounds real nice, for the tutorial to make its way into this project we need the jmd files so I wanted to just take those and concatenate them + remove some fluff:) But happy to colab on the plotting-stuff though!

Also one thing that I've found a bit annoying when doing sampling in Pluto notebooks: sampling can take a bit of time and so having everything re-run everytime you make a change can be annoying if the model is somewhat non-trivial, e.g. in notebook 5 you're looking at a couple of minutes per run. Just a heads up!

kskyten commented 2 years ago

I would also like to have the updated version in my repo as well. I'll want to polish the notebooks a bit anyway before I try anything with Pluto. Native plots would be nice though.

torfjelde commented 2 years ago

I would also like to have the updated version in my repo as well.

Of course :+1:

torfjelde commented 2 years ago

Also one thing that I've found a bit annoying when doing sampling in Pluto notebooks: sampling can take a bit of time and so having everything re-run everytime you make a change can be annoying if the model is somewhat non-trivial, e.g. in notebook 5 you're looking at a couple of minutes per run. Just a heads up!

I just had a look at the Pluto.jl talk for this year's Juliacon and this is no longer the case :tada: You can apparently disable blocks :exploding_head:

shravanngoswamii commented 1 month ago

@torfjelde Is there any plan to add this?