Researchers at Microsoft Semantic Machines are taking a new approach to conversational AI—modeling dialogues with compositional dataflow graphs. Learn how the framework supports flexible, open-ended conversations, and explore the dataset and leaderboard: https://t.co/nPaR8HOhs8
— Microsoft Research (@MSFTResearch) September 21, 2020
Very nice to see this done! Also, this is clearly a busy research area, so try to think "outside of the box" -- or "outside of the circles" : ) #xformers https://t.co/jM61zjDG8s pic.twitter.com/sclZmVGSkG
— Oriol Vinyals (@OriolVinyalsML) September 17, 2020
I love the @TheOnion and the headlines often give me a good chuckle. Titles / headlines of the satirical website have gotten longer and longer over time. I expect that is the case because most people are like me and never click on the articles. Source: https://t.co/PkAqtWaCt4 pic.twitter.com/a495tvZqT7
— Simon Kuestenmacher (@simongerman600) September 15, 2020
I just finished adding the Funnel Transformer model to 🤗 Transformers, a transformer model leveraging ideas from regular CNNs (pooling the hidden states after a block of n layers), ELECTRA pretraining and Transformer-XL positional attention. https://t.co/EG6wF1GWXm
— Sylvain Gugger (@GuggerSylvain) September 15, 2020
A Comparison of LSTM and BERT for Small Corpus
— Thomas (@evolvingstuff) September 15, 2020
"bidirectional LSTM models can achieve significantly higher results than a BERT model for a small dataset"https://t.co/sZDlJ2GtSt pic.twitter.com/ciNL8bhqQE
OpenAI passed a rigorous A/B test in comparison to human-written copy, with a lift of 7%: https://t.co/dHCLwIlx7X
— Greg Brockman (@gdb) September 9, 2020
You can now use huggingtweets to share your own tweet generator with everyone!
— Boris Dayma (@borisdayma) September 9, 2020
Models are trained with @huggingface and tracked with @weights_biases along with their datasets.
My personal favorites are @karpathy and @fchollet
See models at https://t.co/0lwC5cm0Eu pic.twitter.com/ET4HCqV4ct
Want to generate your daily horoscope?
— Hugging Face (@huggingface) September 5, 2020
AstroGPT 🪐 – by @stevhliu
GPT-2 model fine-tuned on Western zodiac signs
Input today's date on:
🔮 https://t.co/uCUVsmgmE7
We've used reinforcement learning from human feedback to train language models for summarization. The resulting models produce better summaries than 10x larger models trained only with supervised learning: https://t.co/Sk31d1CnTu
— OpenAI (@OpenAI) September 4, 2020
Want speedy transformers models w/o a GPU?! 🧐
— Hugging Face (@huggingface) September 1, 2020
Starting with transformers v3.1.0 your models can now run at the speed of light on commodity CPUs thanks to ONNX Runtime quantization!🚀. Check out our 2nd blog post with ONNX Runtime on the subject! 🔥https://t.co/OzPZ8y6XuW
Transformers v3.1.0 is out, first pypi release with 💫 PEGASUS, DPR, mBART 💫
— Hugging Face (@huggingface) September 1, 2020
📖 New & simpler docs and tutorials
🎤 Dialogue & zero-shot pipelines
⭐️ New encoder-decoder architectures: Bert2GPT2, Roberta2Roberta, Longformer2Roberta, ...
📕 Named outputs: pic.twitter.com/T3WzCt8UmY
Microsoft researchers propose a new natural language processing paradigm—one in which models are pretrained from scratch entirely within a specialized domain. Learn about the new BLURB benchmark, leaderboard, and state-of-the-art model for biomedical NLP: https://t.co/xHX7Z26pSa
— Microsoft Research (@MSFTResearch) August 31, 2020
