If you want to get all of them just dount use iloc. If you want to get the first for each id, group by and aggregate with a lambda that filters and takes the first instance

Me da gusto saber que no soy la única persona a la que el iphone le autocorrige "voy" como "voyager" aunque esté desactivado el autocorrect.

I know there's no DMs here but you can find contact info on my github too github.com/chnnxyz

You're welcome. Feel free to hit me up any time with questions on R or Python (specially python, my R is kind of rusty lol) and I will most likely be able to help.

filter(): add as many boolean rules as you want. commas work as AND statements
group_by(): pick one or more variables to group, then operate on them with summarise() or agg

I have some notes but they are in spanish but here is a summary.
select(): grab only the columns passed as arguments or removed if passed as negative
mutate(): create a new column, it can be a function of other columns (see lapply, etc)

also might be too late for the reply since tweet was two days ago but
d1 = df[df['col_a']=='event_a'].iloc[0]['date']
d2 = df[df['col_b']=='event_b'].iloc[0]['date']
then just make d1 and d2 into datetimes if they are strings and subtract or use timedelta.

While I think R is absolute trash in terms of complexity and memory management. I vouch for tidyverse against pandas any single day. It is indeed less intuitive in the beginning because nobody is used to the functional programming paradigm, but I take %>% over chaining methods any day.

And frankly, if you have a plausible idea where to even begin on the problem “how do we teach a LLM object permanence” then congratulations, you are now the most brilliant AI researcher on the planet.

Also it seems that since 2.13 we can just pip install tensorflow github.com/tensorflow/t...

Tensorflow-macos has more recent updates on pip, for starters, and will install the metal api on apple silicon. Dont quote me on this last part because tgere's zero documentation bur i think it installs and manages deps for apple silicon and intel accordingly

tf-metal came a bit after precompiled binaries in conda-forge. also, it is recommended to instead pip install tensorflow-macos