We remain committed to our partnership with OpenAI and have confidence in our product roadmap, our ability to continue to innovate with everything we announced at Microsoft Ignite, and in continuing to support our customers and partners. We look forward to getting to know Emmett

I asked ChatGPT to rewrite Bohemian Rhapsody to be about the life of a postdoc, and the output was flawless:

😍 Lama Cleaner: A free and open-source inpainting tool powered by SOTA AI model 🎁 Completely Free and Open-Source • Fully self-hosted • Multiple SOTA AI models • Classical image inpainting algorithm • CPU & GPU • Various inpainting strategy • Run as a Desktop APP

It's that time of the year again 👀! https://t.co/AJlWUecwNk Everything to know about the past year of happenings in #AI 🎁. My favorite graph: a showcase of how cross-modal #transformers have become 👉

Cool to see @ml6team on @huggingface! https://t.co/RqFO67oR5L

Note: if the interactive version is down for some reason, you can check out the Medium version:

#Transformers 🤖 for #summarization are sometimes prone to hallucinations 😵‍💫. Our #NLP team investigated some post-processing steps in an interactive blogpost on @huggingface 👉 https://t.co/PSWr5WzGky

Machine Learning Operations (MLOps): Overview, Definition, and Architecture https://t.co/2MHAoBjfRN

https://t.co/JckLS4igJ4

For those playing along at home, here's a "AI is sentient!" argument bingo card.

🤩 LayoutLMv3 by @Microsoft is now available @huggingface! 🔥 The model replaces the CNN backbone of its predecessor with (much simpler) patch embeddings à la ViT. 🚀 SOTA performance on all document AI benchmarks, both image-only and text+image! (1/2)

💡 Adapting semantic search to a new domain with unlabeled data 🔍? Using #GPL from @UKPLab, we demonstrate how to do this on non-English datasets 🇩🇪 . 🔗:

Introducing the biggest change to https://t.co/dKZjob5Hcx since its inception 🔥 The Community Tab 🥳 read more and discuss: https://t.co/SI07ac47Z1

📝 I wrote an #NLP blogpost detailing how we optimized our Dutch #GPT2 model to reduce our Google Cloud bill by a factor of 2.4 🤑 ! 👉 https://t.co/DCU4XJV7Bt

🚨 new model alert 🚨. We're releasing: ✌️ translated Dutch summarization datasets (CNN-nl & XSUM-nl) ➕ a finetuned mBART model for Dutch summarization on @huggingface ! https://t.co/WQjkNsFL2t https://t.co/RUsmTF8ck7

It do be like that with #AI 🙈

When your funding round raises about as much dough as a BERT-Base model has parameters 🤯💵! Big congrats to @huggingface 💪!

Our official @huggingface HuggingMugs (or MuggingFaces 🤔) have arrived ☕️! And already our Rouge scores have improved, our Perplexities are down and my eyesight is better 🤯

The @amazon MASSIVE dataset is... well... massive 😱! One million realistic, parallel, labeled virtual-assistant utterances in 51 languages 🎙️🌎.

🤗 Trainer now sports --optim adamw_bnb_8bit which activates the 8-bit Adam optimizer https://t.co/X1KdSIZZNC and uses 6 bytes less per param for training. ~1/3 of total memory savings! Huge thanks to @ManuelCiosici & @Tim_Dettmers for integrating it! Use transformers@main