RadimRehurek's profile picture. Founder and CTO @PII_tools. Applied #ML, #NLP, #IR. 💘 history and beginnings. PhD in AI. Creator @gensim_py. Life & travel in East Asia.

Radim Řehůřek

@RadimRehurek

Founder and CTO @PII_tools. Applied #ML, #NLP, #IR. 💘 history and beginnings. PhD in AI. Creator @gensim_py. Life & travel in East Asia.

Pinned

Scoping and research analysis finished, starting R&D on real data.


The most disgusting publication to cite my @gensim_py (yet)? 🤮 peerj.com/articles/cs-11…

RadimRehurek's tweet image. The most disgusting publication to cite my @gensim_py (yet)? 🤮

peerj.com/articles/cs-11…

BLAS-level CPU Performance in 100 Lines of C: cs.stanford.edu/people/shadjis…

RadimRehurek's tweet image. BLAS-level CPU Performance in 100 Lines of C:
cs.stanford.edu/people/shadjis…

So. Promo code MALICE (from @michaelmalice podcast) at @IPVanish gets me $3.75/mo. But if I go straight to ipvanish.com and ignore Michael's "special offer", I get $3.20/mo 🤔 I don't mind the $$$ but what's the logic here?


Fed up with large companies abusing our #privacy? The wonderful folks at @NOYBeu are fighting them and need your help! Fulltime #webdev please respond here: noyb.eu/sites/default/…

RadimRehurek's tweet image. Fed up with large companies abusing our #privacy? The wonderful folks at @NOYBeu
are fighting them and need your help!

Fulltime #webdev please respond here:
noyb.eu/sites/default/…

Radim Řehůřek reposted

Summer release of Python #smart_open v5.2.0: Faster Azure connections & a new GCS parameter & several fixes 🌻 github.com/RaRe-Technolog…


Oh yes, #Gensim now accepts sponsors :) If your company / job relies on my #opensource work, please consider supporting me to sustain it 🙏 github.com/sponsors/piskv…

Gensim got its first Github sponsor! ❤️ Thank you @WiLabsCom – check them out 🤗 radimrehurek.com/gensim/people.… #opensource #sustainability



Radim Řehůřek reposted

Gensim 4.0 is finally out 🤗 Faster, leaner, better. Full Changelog + Migration notes here: github.com/RaRe-Technolog… Big thanks to all who participated in beta + RC releases.

gensim_py's tweet image. Gensim 4.0 is finally out 🤗 Faster, leaner, better.

Full Changelog + Migration notes here:
github.com/RaRe-Technolog…

Big thanks to all who participated in beta + RC releases.

Radim Řehůřek reposted

💥 Gensim 4.0.0 RC1 (release candidate) is out! 💥 Your last chance to check everything works for you & report issues, before the full 4.0 release tomorrow 🤗 $ pip install --pre --upgrade gensim

gensim_py's tweet image. 💥 Gensim 4.0.0 RC1 (release candidate) is out! 💥

Your last chance to check everything works for you & report issues, before the full 4.0 release tomorrow 🤗

$ pip install --pre --upgrade gensim

Radim Řehůřek reposted

🧵Practical named entity recognizer by @RadimRehurek' team. Somewhat unsurprisingly they use good old convnets. Why this old ... you would ask and no shiny transformers. Well, things have to run very quickly in production.


Q: Why build an in-house #NER engine, in 2021? Why not use #opensource? A: Performance + accuracy + flexibility. Definitely worth it 👌 (if you know what you're doing)


Radim Řehůřek reposted

Winter release of Python #smart_open v4.1.2: github.com/RaRe-Technolog… More options for S3, codecs fixes, faster reads. Enjoy! ☃️


I forgot what the point of that wrapper was, but its execution wasn't great and we removed it completely in Gensim 4.0: github.com/RaRe-Technolog…

Why @gensim_py don't fit their sklearn_api data type for input to the same type sklearn transformers use?? Isn't the whole point of it to fit with sklearns modules?



Radim Řehůřek reposted

#Gensim 4.0 is coming! 💥💥💥 Massive optimizations, general cleanup, new website 💥💥💥 Want to help? Install beta and let us know how it went: $ pip install --pre --upgrade gensim github.com/RaRe-Technolog…

gensim_py's tweet image. #Gensim 4.0 is coming! 💥💥💥 Massive optimizations, general cleanup, new website 💥💥💥

Want to help? Install beta and let us know how it went:

$ pip install --pre --upgrade gensim

github.com/RaRe-Technolog…

A proposal to make CPython (the "standard" #Python) 5x faster 🍽️ mail.python.org/archives/list/…


Deep down, it's always about low-rank matrix decompositions. "Rethinking Attention with Performers" blog post by @GoogleAI ai.googleblog.com/2020/10/rethin…

RadimRehurek's tweet image. Deep down, it's always about low-rank matrix decompositions.

"Rethinking Attention with Performers" blog post by @GoogleAI 

ai.googleblog.com/2020/10/rethin…

This made my day. God bless. #NLProc

RadimRehurek's tweet image. This made my day. God bless. #NLProc

Radim Řehůřek reposted

Released #smart_open 2.2.0: $ pip install smart_open[aws] (dependencies like S3, GCS etc are now optional, for a leaner install) Also, S3 performance improvements. Enjoy! 🌞 github.com/RaRe-Technolog…


Loading...

Something went wrong.


Something went wrong.