Video yükleniyor...
Video Yüklenemedi
🤗🤗🤗introducing Hugging Science -- the home of AI for science 🤗🤗🤗 open models and datasets are the powerhouse of science (see the PDB), but finding the models and data you actually need for your breakthrough is hard af you shouldn't need to scrape arxiv, own your own wetlab, fight... show more
226,376 görüntüleme • 5 ay önce •via X (Twitter)
50 Yorum
This is exactly the move. Discoverability of real scientific data + models on the Hub has been the missing piece for years. OpenMed's 1,000+ medical models all sit on top of @huggingface infra. becoming its own destination is overdue and welcome. 👏
@lvwerra Nice, more incoming from our side
@lvwerra OOOooooOOOo say more?
@lvwerra Ahaha more in counting!!
Great resource! Hope it expands beyond biology to physics (i.e. Materials discovery).
Yeah, tons and tons of materials data and models -- not to mention multiple benchmarks!
Yaaaaaaay 🚀🚀🚀🚀
Excellent initiative thank you @cgeorgiaw !
😊😊😊
Dang, that's incredible! I see potential for expansion. Do you envision this to eventually grow into a science data repository for *everything*?
We want it to be comprehensive but not overwhelming. A huge issue with scientific data is that you have tons and tons of random stuff where you're not sure what it is and where it came from. We are trying to mitigate that issue by putting forward really high quality, well-structured datasets that you can build with this instant.
Makes sense. Great work, that's going to help a lot of researchers.
What do you think is the best large language model for molecular biology questions and projects?
nice! I don't see and CFD datasets in the physics or engineering sections?
Been thinking about including stuff like this: But would love more suggestions. In general trying to include things that are sufficiently well-documented that people could build without a huge deep dive.
There all have dataset cards to some extent or another have a few more in there too
Epic will add TYSM!!!
Congrats on the launch. Can I pitch a challenge and what does standing one up look like? The one I'd love to see: pre-symptomatic Parkinson's detection from voice. Biomarkers predict PD 3-5 years early. Consented speech corpus + biomarker model + years-to-symptom leaderboard.
Yeah, 100%! We have a blog on the site about how to set up a challenge, but it’s pretty straightforward. You basically need a frontend leaderboard and an eval metric(s) that you can calculate from what people submit. You can also clone the set-ups of previous challenges that we’ve done. Once you’ve got it, reach out!
@prasithg I have the same question. Can you point me to the blogpost please?
@prasithg This blog is very how-to oriented, but if you feed your agent of choice with this blog, your idea, and an eval metric it should be able to set up a challenge without an issue.
I have such immense respect for you guys this is HUGE
Hey @atranscendedman check this out
Congratulations!
@ayirpelle Exciting times ahead 😁
This is just great, Hugging Science, I just love it. I have many suggestions eg bacterial datasets downloaded paper by paper that have been accumulated on various hard drives. Have already spotted databases I want to use. Thank you!
I have been trying to upload models and datasets for a while now ....now huge upgrade
Science deserves its own Spotify Wrapped moment for datasets.
😂😂😂
Cool stuff appears on X everyday
🤗x🧬 = ✅
What's the typical workflow you're envisioning? Like researcher searches "protein folding transformers" and gets ranked models with actual performance metrics instead of digging through 50 papers to find which ones have usable code?
That would be sick -- need a leaderboard for that. Even simpler workflow is "I'm trying to train a DNA model, what is the set of high-quality, known-provenance datasets I can use off the bat?" so you can go from idea to MVP way, way faster
科学のための人工知能のホーム、Hugging Scienceを紹介します。開かれたモデルとデータセットは科学の原動力です。しかしながら、実際にあなたのブレークスルーに必要なモデルやデータを見つけるのは本当に難しいですよね。しかし、私たちがそれを変えています。すべての最高の科学を一か所に集めました。さまざまなデータが準備されており、トレーニングが可能です。さらに、ドメイン、タスク、キーワードでフィルタリングや検索もできます。科学の進展に貢献しましょう！
@_lewtun > 11TB of PDEs Well there goes sleep and weekends.
This is awesome! I'm gonna save this for superlab :D
@ClementDelangue did you approve this?
i need DFT and NBO analysis on molecules in aqueous solution!
Woah
honestly the discovery problem is real, my main question is how did you determine "best" vs just having everything - because hf already has a lot of this stuff and the filtering is the actual pain point. like is this curated or just aggregated
curated! we selected for extremely well-documented, extensive, known-provenance datasets and models to minimize the "what even is this???" feeling. if you have any more suggestions how we could do that better, would love that (or just like specific pains too)
stop building stellarators when you can just scrape the data and bill it to your investors honest truth, most labs are just
Congrats!
"Open models and datasets are crucial, but sometimes scraping ArXiv or other sources is necessary for research. It's not about owning everything, but rather having access to what's out there, even if it means some extra legwork."
@Quetzally_Med
Hugging face is so convenient
Long live humanity!
Thank you for this gift to the community 🙏
AWESOME NEW AGE KAGGLE
@_akhaliq What can I build with this? Tell me!!
![There is a beautiful story that just happened in AI so let me share it for a lighter tone weekend post among all the doom stories in our AI field this week. It’s a story of people on three continents building and sharing in the open a new small efficient and state-of-the-art AI model. It started a couple of months ago when a new team in the AI scene released their first model from their headquarters in Paris (France): Mistral 7B. Impressive model, small and very strong performances in the benchmarks, better than all previous models of this size. And open source! So you could build on top of it. Lewis in Bern (Switzerland) and Ed (in Lyon, in the South of France) both from the H4 team, a team of researchers in model fine-tuning and alignment were talking about it over a coffee, in one of these gatherings that often happen at Hugging Face to break the distance between people (literal distance as HF is a remote company). What about fine-tuning it using this new DPO method that a research team from Stanford in California just posted on Arxiv, says one? Hey, that’s a great idea, replies the other. We've just build a great code base (with Nathan, Nazneen, Costa, Younes and all the H4 team and TRL community) let's use it! The next day they start diving in the datasets openly shared on the HF hub and stumble upon two interesting large and good quality fine-tuning datasets recently open-sourced by OpenBMB, a Chinese team from Tsinghua: UltraFeedback and UltraChat. A few rounds of training experiments confirm the intuition, the resulting model is super strong, by far the strongest they have ever seen in their benchmarks from Berkeley and Stanford (LMSYS and Alpaca). Join Clementine, the big boss of the open evaluation leaderboard. Her deep dive into the model capabilities confirms the results: impressive performance. But the H4 team also hosts a famous faculty member, Pr. Sasha Rush, Associate Professor at Cornell University in his daytime, hacker at HF in his nighttime. Joining the conversation, he proposes to quickly draft a research paper to organize and share all the details with the community. A few days later, the model, called Zephyr (a wind like Mistral), paper, and all details are shared with the world. Quickly other companies, everywhere in the world starts to use it. LlamaIndex, a famous data framework and community, shares how the model blew their expectations on real-life use-case benchmarks, while researchers and practitioners discuss the paper and work on the Hugging Face hub. All this happened in just a few weeks catalyzed by open access to knowledge, models, research, and datasets released all over the world (Europe, California, China) and by the idea that people can build upon one another work in AI to bring real-world value with efficient and open models. Stories like this are numerous everywhere around us and make me really proud of the AI community and see how we can build amazingly useful things together. [the video is just me reading this Friday post hahah]](https://image.24vids.com/tw-1720503998518640703/ext_tw_video_thumb/1720503027273023490/pu/img/VOLi_zPnZlb1nuWB.jpg)
