Shout out to @cloudconvert: took me only minutes to set up a document conversion pipeline with them. The whole experience was super smooth (testing on website, API, docs, payment setup etc). Bonus points: EU only hosting and also from Munich 👍
Matej Svejda retweeted
Having trouble with code quality? Write a lot of tests. And I mean a *lot*. Test every file in isolation. Mock as many imports as possible.
When you're done, your code will still be bad, but now your tests will make sure it's impossible to improve anything in any meaningful way.
This honestly seems like the only really viable approach right now. Not sure about germany starting to control the outbreak though, doesnt seem like it.
I believe the US and many other nations can/should transition to a scalable test/trace/quarantine approach against #COVID19 in the medium term. See detailed discussion here: hunch.net/?p=13762539 .
Matej Svejda retweeted
I believe the US and many other nations can/should transition to a scalable test/trace/quarantine approach against #COVID19 in the medium term. See detailed discussion here: hunch.net/?p=13762539 .
Something that Condens had from the very start that seems to resonate very well with our users.
This Post is from an account that no longer exists. Learn more
Appreciation tweet of the day: @MarijnJH not only develops great software @ProseMirror but also maintains it sooooo well. Thanks so much for that!
Matej Svejda retweeted
"The story of global poverty is more complex than you and Gates have been willing to acknowledge, and the data do not support your rosy narrative about neoliberal globalization." jasonhickel.org/blog/2019/2/…
Matej Svejda retweeted
Trump doesn’t understand climate change and @ronnychieng is sick of explaining this s**t. on.cc.com/2Wy9BRk
Very cool visualization of what attention doesn in BERT.
What goes on in the mind of BERT? Using an interactive visualization tool, we uncovered some surprisingly intuitive patterns.
Article: medium.com/@JesseVig/deconst…
Colab: colab.research.google.com/dr…
Github: github.com/jessevig/bertviz
Play with it and share what you find! #nlproc
Really well made and fun to read article about the current state-of-the-art in NLP transfer learning. Also while you are there check out the other articles written by @jalammar
This Post is from an account that no longer exists. Learn more
No-one writes blog posts like @pabloariasal A joy to read - as always!
This quoted post is unavailable.
Pretty good summary by @nytimes of why it's such an exciting time for NLP right now: nytimes.com/2018/11/18/techn…
Matej Svejda retweeted
Have been experimenting with an unsupervised pre-training technique for @spacy_io, similar to ULMFit/Elmo/BERT etc. Still very early, but pushing to spacy-nightly soon so you can check it out! 🚀💫
github.com/explosion/spaCy/p…
Matej Svejda retweeted
The multilingual BERT model is out now (earlier than anticipated). It covers 102 languages and features an extensive README motivating certain preprocessing and modelling choices.
github.com/google-research/b…