Come for the machine learning stay for the commentary on everything else.
Based in Berlin. Doing statistics with style and flair.
| Pronouns | he/him |
I have not! Can you tell me more?
So like I'm probably going to be deported soon from Germany if I can't find a job here. If you need a software developer who does a lot of machine learning, DM me
Ok it still crashes occasionally, but your patch fixed my system. Thanks so much for this! I really hope the KDE devs upstream it
Hey @thaodan I noticed you were having the same issue as me getting i3 to run under KDE Plasma 6.5
How do I build your forked repo to get your fix running?
A major reason lots of people don't like eDSLs is that many of them don't have good debugging and other tooling.
And it makes sense, you make an eDSL so you can lean on the host language. But, many of them end up being a miserable user experience since use that as license to not add lots of niceties people expect from languages these days.
Only if that language is Python! The broadcasting rules make it hard to do efficient autodiff as Adrian Seyboldt goes into in this thread:
https://discourse.pymc.io/t/user-experience-python-vs-r-pymc-vs-stan-pytensor-vs-jax/16426/28
The numpy broadcasting rules are actually a bit of a pain to work with in pytensor too. If you want to follow it strictly, you can’t tell until runtime what for instance a multiplication of two arrays will actually do. Depending on if a shape is one, it might broadcast, error out, or do elementwise multiplication. But which one happens has implications for the derivatives, because those won’t be the same in those cases. If I could go back in time, I’d love to tell the numpy devs to not use a d...
