Christoph Schwienheer

I make stuff with code and data.

[Christoph Schwienheer @chs_sch]

@francoisfleuret Zuversicht

[Christoph Schwienheer @chs_sch]

Draw some #Wordle clues from a @spacy_io vocab:
https://t.co/yi76gW2GbS

[Christoph Schwienheer @chs_sch]

@abhi1thakur On it just right now and also just got this meta-ish example in. @explosion_ai @spacy_io

[Nils Reimers @Nils_Reimers]

Big, Bigger, Biggest🏋️
Join me to train the best Sentence Embedding Model ever on 1B training pairs

At @huggingface the JAX/Flax community event is upcoming with @GoogleAI providing free TPUv3 VMs.

Join me and train the best Dense Embedding Model ever
https://t.co/NFNEGots2v

[Ines Montani @_inesmontani]

NEW BLOG POST: Applied NLP Thinking – a mindset for translating business problems into ML solutions.

TL;DR
🧠 Think about the "what", not just the "how"
📈 Check utility, not just accuracy
🚀 Make prototype to production smooth
📚 Linguistics can help!

https://t.co/gbdp5BoeAE

[Christoph Schwienheer @chs_sch]

Friends just started this crowdfunding, recommended (if you like licorice): Lakki - Lakritzschnaps aus Hamburg https://t.co/BM3qe3xhNC via @startnext

[Pierre Lison @plison2]

Proud to present *skweak*, a new Python library to apply weak supervision to NLP tasks: https://t.co/2RGJaXIBwJ Don't annotate texts by hand - use skweak instead! skweak is easy to use and tightly integrated with @spacy_io . Joint work with @jeremy_nlp & @AliaksandrHubin

[Nils Reimers @Nils_Reimers]

🚨 Model Alert - Dense Information Retrieval

Want to achieve nearly the same performance as the ANCE model, but 5x faster + 2x smaller embedding size?

Check out our new SBERT models for MSMARCO using MiniLM:
https://t.co/eWNH82kwKA
https://t.co/tFCrc4Kmz7

[GGG @OptionSpike]

READ THIS - RETWEET IF YOU CAN 🦍✊💎🍋 $GME $AMC $NASDAQ $DJIA $SPY $SEC $STOCKS

[Christoph Schwienheer @chs_sch]

@Hetzner_Online https://t.co/QZsLUDJLxo

[CuPy @CuPy_Team]

We have found a PyPI package named "cupy-cuda112" which was NOT uploaded by the CuPy project. Please do not install it. We are contacting @PyPI to address this issue. https://t.co/6MpMXR5i02

[Samuel Colvin @samuelcolvin]

Pydantic just overtook marshmallow to become the most popular standalone python data validation library but github stars. https://t.co/90HDkGcHMP

[Christoph Schwienheer @chs_sch]

Idea: An AI that is able to create a working CUDA Docker container on the first try. 😠

[Christoph Schwienheer @chs_sch]

I ran already a lot of experiments with the nightly and was amazed by the usage approach and great engineering/desing decisions. Looking forward to check out the release!

[Christoph Schwienheer @chs_sch]

@343max @ewyler @mspro @ennopark https://t.co/TeSsyC3qQh

[Christoph Schwienheer @chs_sch]

@honnibal @spacy_io @Apple Thanks. Yeah right, I was just wondering where the wrapping / compability happens.