We are releasing mT5: A massively-multilingual version of T5 that supports over 💯 languages! mT5 was pre-trained on a multilingual version of C4 and achieves SoTA on many cross-lingual NLP tasks.
— Adam Roberts (@ada_rob) October 23, 2020
📜Pre-print: https://t.co/lfQd9CeF7b
💾Code/models: https://t.co/1MauV6etQY pic.twitter.com/LUJIFaMxF5
This looks like quite an amazing Python Profilerhttps://t.co/KD3v70rxsQ pic.twitter.com/ja900vG08f
— Jeremy Howard (@jeremyphoward) October 22, 2020
PyTorch Lightning 1.0.0 is now available. This is the final stable API to train and deploy models at scale, without the boilerplate. Read more about this release below: https://t.co/9zozqQWOCL
— PyTorch (@PyTorch) October 21, 2020
Data science team question! What are people's favorite ways for a team to store completed analyses? A shared folder? A wiki page? A tool specifically for this like Airbnb's? https://t.co/reykO6CCvH
— Jacqueline Nolis (@skyetetra) October 20, 2020
Goal is to have a nice shared place where you can keep results forever.
🚨New model in town!!🚨
— Hugging Face (@huggingface) October 20, 2020
ProphetNet is a new pretrained seq2seq model. We just added the model in 🤗 Transformers! Both English and multilingual pretrained models are available from 🤗 Model Hub!
👉 Doc: https://t.co/k5UKScrgO9@MSFTResearch @QiWeizhen @gyynlpanda @ruofeibruce pic.twitter.com/06UEJdBhwQ
This is true. I've started a few talks about ML progress with an interactive demo using https://t.co/8D6k1uEpeG, and it's amazing how quickly you can build intuitions in a large audience in just a couple of minutes with these visualizations! https://t.co/2LlSQLD9gP
— Jeff Dean (@🏡) (@JeffDean) October 17, 2020
I've been playing with @TheRealDAGsHub (https://t.co/WyrVQgcVvM) and @DVCorg and it's pretty neat
— Julien Chaumond (@julien_c) October 16, 2020
My pipeline is https://t.co/DRHWqGzj8X
This feels like a "structured Weights&Biases", if it makes sense.
Anyone tried it yet? pic.twitter.com/hC1j757DJC
`fastdoc`, the system that @GuggerSylvain and I used to write our beautiful @OReillyMedia book, is now available for you to use (if you're crazy enough to think writing a book is a good idea...)https://t.co/g4La0cQLp4 pic.twitter.com/D8arAFjdoi
— Jeremy Howard (@jeremyphoward) October 12, 2020
This new feature of fastcore is very cool! ✨✨Runtime✨✨ type checking in Python 😍. https://t.co/ls08Kg4Vr8 pic.twitter.com/5NHIzkXERt
— Hamel Husain (@HamelHusain) October 12, 2020
Ceyda Cinarel (@ceyda_cinarel) shared a neat demo of a custom inference widget for token classification models 🏷, built on top of the:
— Julien Chaumond (@julien_c) October 9, 2020
- @huggingface Inference API
- Displacy from @spacy_io
- @streamlit
🔥 Demo: https://t.co/Rlk5gb9rO0
🔥 Blogpost: https://t.co/rKwlrFeMj2
An open-source PyTorch library for Graph Transformer Networks.
— Yann LeCun (@ylecun) October 8, 2020
Allows you to build DL systems that perform operations on weighted graphs and bakcpropagate gradients through these operations.
(this is different from graph neural nets).
From FAIR.https://t.co/e6iMzPMLhB
If you're reviewing a paper for the second+ time, Draftable is a lifesaver: https://t.co/3xvEPi9IC5. It allows you to easily and quickly see changes made to a manuscript.
— Tatyana Deryugina (@TDeryugina) October 8, 2020
I've tweeted about this tool before, but worth re-iterating.
