1️⃣5️⃣ things you probably didn't know about Hugging Face!
That's a 15min talk I'll give today in SF - and 🚨 SPOILER ALERT 🚨 I have it here for you as a sneak preview 🧵
Want to see it live? Register 👇
eventbrite.com/e/future-tech-…
That's a 15min talk I'll give today in SF - and 🚨 SPOILER ALERT 🚨 I have it here for you as a sneak preview 🧵
Want to see it live? Register 👇
eventbrite.com/e/future-tech-…
1️⃣ Democratize Good Machine Learning
Everything we do at 🤗 rolls up to our mission to democratize good machine learning, and we do this through open source, community-driven, and ethics-first machine learning.
Everything we do at 🤗 rolls up to our mission to democratize good machine learning, and we do this through open source, community-driven, and ethics-first machine learning.
2️⃣ Speaking of ethics-first, let's start with that.
We want to operationalize ethics in AI, that means creating tools for the community to build ML with ethics-first principles.
Find them all on this space by @mmitchell_ai @YJernite @NimaBoscarino 🛠🔬
huggingface.co/spaces/society…
We want to operationalize ethics in AI, that means creating tools for the community to build ML with ethics-first principles.
Find them all on this space by @mmitchell_ai @YJernite @NimaBoscarino 🛠🔬
huggingface.co/spaces/society…
3️⃣ Hugging Face is one of the top open source organizations in the world.
Maybe you know Transformers and Datasets, but do you know Diffusers, Accelerate, Optimum, PEFT, SetFit?
(🚀 now #36 since I wrote this last week with 185k ⭐️ on GitHub - see gitstar-ranking.com/organizations)
Maybe you know Transformers and Datasets, but do you know Diffusers, Accelerate, Optimum, PEFT, SetFit?
(🚀 now #36 since I wrote this last week with 185k ⭐️ on GitHub - see gitstar-ranking.com/organizations)
4️⃣ Diffusers is growing faster than Transformers did
Transformers made it super easy to build with the latest models and switch ML framework, ML task or model architecture.
Diffusers is doing the same for diffusion models for image and audio generation. Follow @diffuserslib !
Transformers made it super easy to build with the latest models and switch ML framework, ML task or model architecture.
Diffusers is doing the same for diffusion models for image and audio generation. Follow @diffuserslib !
5️⃣ You can find >150k free ML models on the 🤗 Hub.
It's the largest source of state of the art models, and you can find anything from NLP to vision, audio, RL, transformers and models from 35+ libs.
All free and accessible thanks to the 🤗🌎 community!
huggingface.co/models
It's the largest source of state of the art models, and you can find anything from NLP to vision, audio, RL, transformers and models from 35+ libs.
All free and accessible thanks to the 🤗🌎 community!
huggingface.co/models
6️⃣ And Hugging Face hosts even more private models!
When you create a model, dataset or Space repo, you can set it public or private - which is great to collaborate with your team!
Over 15,000 organizations use 🤗 to build ML into their apps.
When you create a model, dataset or Space repo, you can set it public or private - which is great to collaborate with your team!
Over 15,000 organizations use 🤗 to build ML into their apps.
7️⃣ HF users download over 1 million models a day 🤯
And our stats are open source too!
Check out model download trends at huggingface.co/spaces/hugging…
(BTW pro tip - @lizallendorf added a feature for repo owners to see lifetime downloads - find it in repo settings)
And our stats are open source too!
Check out model download trends at huggingface.co/spaces/hugging…
(BTW pro tip - @lizallendorf added a feature for repo owners to see lifetime downloads - find it in repo settings)
8️⃣ How we make money 🤑
... is a little known fact 😬
HF makes money by selling:
🏭 Compute services to train, demo and deploy models for production: hf.co/pricing
🧑💻 Expert support to help companies build with open source ML: hf.co/support
... is a little known fact 😬
HF makes money by selling:
🏭 Compute services to train, demo and deploy models for production: hf.co/pricing
🧑💻 Expert support to help companies build with open source ML: hf.co/support
9️⃣ Developers: get started with /tasks
Want to build ML features without combing through 150k models?
Head over to hf.co/tasks led by @mervenoyann
Pick a task like text summarization, it will recommend the best models, and you can even deploy your own API!
Want to build ML features without combing through 150k models?
Head over to hf.co/tasks led by @mervenoyann
Pick a task like text summarization, it will recommend the best models, and you can even deploy your own API!
1️⃣0️⃣ Expect open source LLMs from Hugging Face!
We are committed to providing open source alternatives to proprietary models. We built a supercomputer cluster for open source research - check out BigCode hf.co/bigcode and follow @SanhEstPasMoi for Flamingo repro!
We are committed to providing open source alternatives to proprietary models. We built a supercomputer cluster for open source research - check out BigCode hf.co/bigcode and follow @SanhEstPasMoi for Flamingo repro!
1️⃣1️⃣ BLOOMZ is the largest open source instruction LLM
Many know about 🌸 BigScience/BLOOM, but few know BLOOMZ, a fully open, instruction tuned, 176B parameter LLM - best multilingual basis for open RLHF projects!
hf.co/bigscience/blo…
Many know about 🌸 BigScience/BLOOM, but few know BLOOMZ, a fully open, instruction tuned, 176B parameter LLM - best multilingual basis for open RLHF projects!
hf.co/bigscience/blo…
1️⃣2️⃣ ControlNet is 🔥 for image generation!
You know Stable Diffusion - but do you know ControlNet?
Use doodles, contours, depth, human pose as guides to get exactly the image you want.
And with @diffuserslib - it's wicked fast!
Play with it at huggingface.co/spaces/hysts/C…
You know Stable Diffusion - but do you know ControlNet?
Use doodles, contours, depth, human pose as guides to get exactly the image you want.
And with @diffuserslib - it's wicked fast!
Play with it at huggingface.co/spaces/hysts/C…
1️⃣3️⃣ There's a (Mac) app for Diffusers!
We offer many ways to experiment with models like Stable Diffusion, but did you know you can play with the models in a few clicks and get your Mac to create the images for you?
Get it ➡️ apps.apple.com/app/diffusers/…
(needs M1 or later chip)
We offer many ways to experiment with models like Stable Diffusion, but did you know you can play with the models in a few clicks and get your Mac to create the images for you?
Get it ➡️ apps.apple.com/app/diffusers/…
(needs M1 or later chip)
1️⃣4️⃣ Hugging Face offers a free course on Generative AI!
Learn how to build apps that generate images and audio, from theory to practice 🎓🚀
github.com/huggingface/di…
Learn how to build apps that generate images and audio, from theory to practice 🎓🚀
github.com/huggingface/di…
1️⃣5️⃣ Add ML into Web apps with huggingface.js 🔥
2/3 of devs build in JavaScript - now they can add ML into their apps with our new open source client library!
Serve predictions from any Hugging Face models directly from a webpage! 🤯
github.com/huggingface/hu…
2/3 of devs build in JavaScript - now they can add ML into their apps with our new open source client library!
Serve predictions from any Hugging Face models directly from a webpage! 🤯
github.com/huggingface/hu…
Now I've open sourced my talk, will anyone show up tonight at Buzzworks in SF? 🤣
All the above is releases and milestones over the last few weeks. If this was useful to you, please let me know with a retweet so I open source my next talk! 🙋
All the above is releases and milestones over the last few weeks. If this was useful to you, please let me know with a retweet so I open source my next talk! 🙋
• • •
Missing some Tweet in this thread? You can try to
force a refresh
