Closed graeme-a-stewart closed 10 months ago
PR Preview Action v1.4.4
:---:
:rocket: Deployed preview to https://JuliaHEP.github.io/JuliaHEP-2023/pr-preview/pr-36/
on branch gh-pages
at 2023-11-10 10:03 UTC
If we're fixing typos, then in dataframes intro, "manegable" -> "manageable".
(Thanks for spotting my "exponentation" in return!)
Thanks! For some reason my code spell checked doesn't run in markdown cells.
Yah, the point is that Jupyterbook moans as it compiles all the ipynb
files and then the solutions notebook crashes as Zygote isn't there.
Idk if we need Zygote related content. Zygote may phase into "not recommend" in a couple of years
Idk if we need Zygote related content. Zygote may phase into "not recommend" in a couple of years
Well, it's just the "very advanced optional" bit of the exercise, as an excuse to flag the very advanced auto differentiation ecosystem in Julia. Do you have a suggested replacement out of the many alternatives? ReverseDiff/ForwardDiff? Enzyme? (which I would have used over Zygote, but my impression was that there's still issues with Enzyme?)
Unless we have some specific HEP AD thing, I'd suggest not go too deep into something just for the sake of it
To clarify, I mean "very advanced" as in "it's at a state of mature development", not "I think it's super complicated to understand". The exercise literally just touches on it at the end so people know that autodiff stuff exists (and where to find it) in Julia, and it only uses Zygote because it's both easy to use in this case, and I can guarantee it will work in 1.9.3 (unlike Enzyme!)
What if we make clear Zygote is just an example and encourage readers to explore https://juliadiff.org/ for example?
I would argue the text already does that! ("There are many automatic differentiation packages available for Julia - such as Zygote
- listed at Julia Diff, which we could use to find f′
directly and efficiently.")
I was just thinking not making Zygote sounds like the go-to flagship because to some people it's beyond salvation but I guess this is fine!
Sure, I guess the problem is that it's hard to recommend the best replacement (since Diffractor and Enzyme both have their champions).
Update of various tutorial aspects:
DataFrames:
EventID, PRI*, Label
)select
transform
missing
values work: types + multiple dispatchOperators and Variables:
Control and Conditional:
General:
Closes #33 Closes #30