What you may have missed from the 🤗 open source community gathering in Paris last week: demos 🕹️

The ~2k folks who attended, while mingling, stumbled upon demos from the community: companies, researchers, artists and tinkerers ✨

A bit more about each of them 👇🧵 1 - Rewyse: a platform that uses ML and NLP AI to provide insights and analysis on product reviews at scale rewyse.ai

We just released Transformers' boldest feature: Transformers Agents.

This removes the barrier of entry to machine learning

Control 100,000+ HF models by talking to Transformers and Diffusers

Fully multimodal agent: text, images, video, audio, docs...🌎

huggingface.co/docs/transform… Create an agent using LLMs (OpenAssistant, StarCoder, OpenAI ...) and start talking to transformers and diffusers

It responds to complex queries and offers a chat mode. Create images using your words, have the agent read the summary of websites out loud, read through a PDF

SAM, the groundbreaking segmentation model from @Meta is now in available in 🤗 Transformers!
What does this mean?

1. One line of code to load it, one line to run it
2. Efficient batching support to generate multiple masks
3. pipeline support for easier usage

More details: 🧵 You can first read more about the model, and learn how to use it on our documentation page: huggingface.co/docs/transform…
Let's check all the features we support below!

It's been an exciting year for 🤗Transformers. We tripled the number of weekly active users over 2022, with over 1M users most weeks now and 300k daily pip installs on average🤯 We doubled the number of architectures (89 to 167🤯) with new models in audio🔊, text📚, vision🖼️, multiple modalities or even time series📈and protein folding🧬

Here are a few highlights in the most used of those new models👇

📊4 challenging speech tasks, 102 spoken languages: can one model solve them all? 🤯

Introducing @GoogleAI's XTREME-S🏂 - the first multilingual speech benchmark that is both diverse, fully accessible, and reproducible!

👉huggingface.co/datasets/googl…

1/9 XTREME-S covers
- automatic speech recognition (ASR),
- speech translation (ST),
- speech classification, and
- speech retrieval.

2/9

[THREAD] Following the public release of Spaces, here is a showcase of a few ones we like. Let’s start with this surprising Draw-to-Search demo by @osanseviero and powered by CLIP. huggingface.co/spaces/osansev… Next, listen to the tacotron2 voice reading you how to bake cookies (in Mandarin) with this Coqui Text-to-Speech demo by @eugene_siow. huggingface.co/spaces/eugenes…

Part 1 of the course focused on text classification, part 2 will focus on all other common NLP tasks. @mervenoyann has made videos to introduce you to each of them!
Let's start with Token Classification (giving a label to some/each word in a sentence):

Then there is question answering: finding the answer to a question in some context.

Transformers can read and write, but how well can they listen and speak 🗣️?

Find out by pitting your models against the SUPERB Challenge 📊!

SUPERB tests pretrained models on a wide range of speech processing tasks & datasets

Submit here 👉: superbbenchmark.org SUPERB aims to accelerate the development of *universal* speech representations 🤯

You can find all the details about the challenge here 👉: superbbenchmark.org/challenge

The results for the JAX/Flax sprint are in🎉. Here are the top-3 projects picked by our awesome jury @wightmanr @nikiparmar09 @ashVaswani @Thom_Wolf
🥇 DALL.E mini hf.co/spaces/flax-co…
🥈 CLIP+NeRF hf.co/spaces/flax-co…
🥉 CLIP for Satellite images hf.co/spaces/sujitpa… Starting today all 🤗 Spaces are publicly viewable 🚀 You can find all the amazing demos created as part of the sprint here 👉 huggingface.co/spaces

We've been getting lots of questions on how to contribute models to🤗Transformers.

Recently we started to publish model-specific recipes on how to do so!

If you want to get better at open-source contributions and want to contribute to🤗Transformers, here is how it works: 👇 1. Watch out for open proposals to add a model here:
github.com/huggingface/tr…

Blog alert: check out the new guest post by Amog Kamsetty and the @raydistributed team on training a Retrieval Augmented Generation Model with Hugging Face and Ray!

huggingface.co/blog/ray-rag The RAG model by @olapiktus @PSH_Lewis and @facebookai colleagues leverages external knowledge sources like Wikipedia to have direct and dynamic access to information at inference time

Part of this process relies on training a retriever to learn how to find that information

1/4. Four NLP tutorials are now available on
@kaggle
! It's now easier than ever to leverage tokenizers and transformer models like BERT, GPT2, RoBERTa, XLNet, DistilBERT,... for your next competition! 💪💪💪! #NLProc #NLP #DataScience #kaggle 2/4. Tokenizers - Training your own tokenizer kaggle.com/funtowiczmo/hu…