Andreas Steiner @AndreasPSteiner
Researching #ComputerVision at #Google using JAX/Flax (https://t.co/Sz1Dg3tKwD). views are my own. Zurich, Switzerland Joined August 2021-
Tweets45
-
Followers559
-
Following121
-
Likes101
Fractals in language?? Fascinating explorations from some colleagues at GDM.
How is next-token prediction capable of such intelligent behavior? I’m very excited to share our work, where we study the fractal structure of language. TLDR: thinking of next-token prediction in language as “word statistics” is a big oversimplification! arxiv.org/abs/2402.01825
Come see us at NeurIPS this afternoon! Michael and CapPa will be presenting the oral 6C, and I'll join them for the poster presentation #214...
Come see us at NeurIPS this afternoon! Michael and CapPa will be presenting the oral 6C, and I'll join them for the poster presentation #214...
Introducing Soft MoE! Sparse MoEs are a popular method for increasing the model size without increasing its cost, but they come with several issues. Soft MoEs avoid them and significantly outperform ViT and different Sparse MoEs on image classification. arxiv.org/abs/2308.00951
NaViT beautifully explained with an animation (which shows both aspect preservation and packing).
NaViT beautifully explained with an animation (which shows both aspect preservation and packing).
NaViT lets you process more images with the same compute – with varying resolutions and aspect sizes! The performance gains translate to OOD datasets with extreme aspect ratios – with the a cropping tuned for OG square ViT, and even more so with a simple "no-prior" resize.
NaViT lets you process more images with the same compute – with varying resolutions and aspect sizes! The performance gains translate to OOD datasets with extreme aspect ratios – with the a cropping tuned for OG square ViT, and even more so with a simple "no-prior" resize. https://t.co/KKBMNALOEy
We went back and compared generative with contrastive image-text pretraining on an equal footing in arxiv.org/abs/2306.07915 Enjoy Lucas's great summary of the main findings:
We went back and compared generative with contrastive image-text pretraining on an equal footing in arxiv.org/abs/2306.07915 Enjoy Lucas's great summary of the main findings:
Computer science that you can touch? If you're in ZRH, come and check out our nuru.nu Seli installation tonight at the D18 Art&Tech Show in Rote Fabrik, to experience the wonderful NCA work from @zzznah @eyvindn @RandazzoEttore @drmichaellevin
Great writeup by @mervenoyann and @RisingSayak about our recent collaboration with HuggingFace for training ControlNet on powerful TPUs – check out the great demos and talks linked in the blog!
Great writeup by @mervenoyann and @RisingSayak about our recent collaboration with HuggingFace for training ControlNet on powerful TPUs – check out the great demos and talks linked in the blog!
Quick summary of our recent work on scaling Vision Transformers - solving stability issues, making training more efficient and cool results: ai.googleblog.com/2023/03/scalin…
To kick-off the JAX diffusers event we are having a series of talks for three days! 🧑🎤👨🎤👩🎤 On 17th of April we will be hosting @AndreasPSteiner from Google Brain, @borisdayma of @craiyonAI and @mmitchell_ai of @huggingface 🔔 Set your github.com/huggingface/co……
You probably know Stable Diffusion by now. But do you know ControlNet? Do you know JAX/Flax? We're planning a hackathon where you will learn about 🤗 diffusers, ControlNet and JAX/Flax. And get some free TPUv4 to fine-tune Stable Diffusion!
You probably know Stable Diffusion by now. But do you know ControlNet? Do you know JAX/Flax? We're planning a hackathon where you will learn about 🤗 diffusers, ControlNet and JAX/Flax. And get some free TPUv4 to fine-tune Stable Diffusion!
A principled approach and lots of good tips to get the best out of your ML models... Highly recommended!
A principled approach and lots of good tips to get the best out of your ML models... Highly recommended!
Tomorrow (Thursday Dec 1st) I’ll present our work “On the Adversarial Robustness of Mixture of Experts” (arxiv.org/abs/2210.10253) at @NeurIPSConf, during the 6th poster session (poster #407). Here’s a quick summary of our work. 🧵
For those interested in a high-level overview, a quick summary of our work on the Pathways Language Image model on Google ResearchBytes. youtube.com/watch?v=K12bYQ…
I'm excited to co-organize an #IJCV special issue on The Promises and Dangers of Large Vision Models with @kaiyangzhou, @liuziwei7, @ChunyuanLi, @kate_saenko_. If you are working on related topics, consider submitting a paper, deadline in 163 days! Link:kaiyangzhou.github.io/assets/cfp_ijc…
With the growing adoption of #AI, mitigating unintended bias is essential. Do fairness properties transfer? How do we debias with arbitrary # of classes/groups? Can debiasing be made interpretable? Below is a short thread about 3 ML fairness papers to appear @ #NeurIPS2022 [1/5]
Lucas Beyer (bl16) @giffmana
56K Followers 443 Following Researcher (Google DeepMind/Brain in Zürich, ex-RWTH Aachen), Gamer, Hacker, Belgian. Mostly gave up trying mastodon as [email protected]Ross Wightman @wightmanr
18K Followers 1K Following Computer Vision @ 🤗. Ex head of Software, Firmware Engineering at a Canadian 🦄. Currently building ML, AI systems or investing in startups that do it better.Jeremy Howard @jeremyphoward
221K Followers 5K Following 🇦🇺 Co-founder: @AnswerDotAI & @FastDotAI ; Hon Professor: @UQSchoolITEE ; Digital Fellow: @StanfordEric Jang @ericjang11
69K Followers 3K Following physical AGI at 1X. Author of "AI is Good for You" https://t.co/eFg4WXhg0pJoan Puigcerver @joapuipe
864 Followers 375 Following Software Engineer in Research at Google DeepMind, Zürich.Sara Hooker @sarahookr
39K Followers 7K Following I lead @CohereForAI. Formerly Research @Google Brain @GoogleDeepmind. ML Efficiency at scale, LLMs, @trustworthy_ml. Changing spaces where breakthroughs happen.rohan anil @_arohan_
12K Followers 2K Following Principal Engineer, @GoogleDeepMind Gemini. prev PaLM-2. Tinkering with optimization and distributed systems. opinions are my own.Basil Mustafa @_basilM
1K Followers 130 Following researching ML @ google brain ZRH | no strong opinions about AI, but very strong opinions about why herbal infusions are awful and should not be called teasJeff Dean (@🏡) @JeffDean
295K Followers 6K Following Chief Scientist, Google DeepMind and Google Research. Co-designer/implementor of things like @TensorFlow, MapReduce, Bigtable, Spanner, Gemini .. (he/him)François Fleuret @francoisfleuret
30K Followers 477 Following Prof. @Unige_en, Adjunct Prof. @EPFL_en, Research Fellow @idiap_ch, co-founder @nc_shape. AI and machine learning since 1994. I like reality.Aakanksha Chowdhery @achowdhery
7K Followers 3K Following LLMs @ Google DeepMind :: PaLM, Gemini // Previously @MSFTResearch, @Stanford, @Princeton // views my own and subject to changeCristian Garcia @cgarciae88
6K Followers 1K Following JAX/Flax at Google DeepMind | Open Source | 🇨🇴Jeremiah Harmsen @JeremiahHarmsen
1K Followers 486 Following Creator of #TensorFlowHub and @TensorFlow Serving. Lead in Google Brain.Neil Houlsby @neilhoulsby
4K Followers 317 Following Professional AI researcher; amateur athlete. Senior Staff RS in the Google Deepmind, Zürich. Attempts triathlons.Matthew Johnson @SingularMattrix
12K Followers 3K Following Researcher at Google Brain. I work on JAX (https://t.co/UGa5tGfinF).Ben Poole @poolio
17K Followers 1K Following research scientist at google brain. phd in neural nonsense from stanford.Andreas Kirsch 🇮�.. @BlackHC
9K Followers 4K Following Past: 🧑🎓 DPhil @AIMS_oxford @ExeterCollegeOx @UniofOxford (4.5yr) 🧙♂️ RE @DeepMind (1yr) 📺 SWE @Google (3yrs) 🎓 @TU_Muenchen 👤 Fellow @nwspkJames Bradbury @jekbradbury
10K Followers 8K Following Compute at @AnthropicAI! Previously JAX, TPUs, and LLMs at Google, MetaMind/@SFResearch, @Stanford Linguistics, @Caixin.Samuel @nblcscd93808
0 Followers 230 Following Musim sejuk berlalu dan musim bunga datang lagi, dan masa mengalir seperti air.Thim7_m254 @Thim7M46706
1 Followers 494 Followingus_Ashley_ @usAshley393605
2 Followers 547 FollowingEva Louise Marie Gabr.. @e681554349
6 Followers 3K FollowingAtulDwivedi964 @dwivedi96424116
0 Followers 33 FollowingJanhavee Shinde @SJanhavee
51 Followers 2K FollowingJuan Hmmm @JuanAH03488233
70 Followers 3K FollowingGrinGina @GrinGina62327
6 Followers 710 FollowingBartosz Cywinski @bartoszcyw
24 Followers 443 FollowingJordan Gong @jordan__gong
41 Followers 2K FollowingCrazyRichBayesians @CrzyRchBayesian
149 Followers 2K Following definitely crazy, not rich, probably bayesianHaoli Yin @HaoliYin
260 Followers 822 Following Incoming RS Intern @datologyai | Multimodal, Data-Centric AI | prev @modern_ai, @VU_Biophotonicspawann k. @pawaniiit
220 Followers 4K Following Prof., PhD, Inria, France, Postdoc KU Leuven, Fraunhofer ITWM, FU Berlin. I like Machine learning and mathematics.Michael Vorburger ⛑.. @vorburger
532 Followers 464 Following So long, 🐦 Twitter... see ya all on #Mastodon! 🦣Shambhavi Sinha @SSinha_154
4 Followers 74 FollowingSimon Batzner @simonbatzner
4K Followers 669 Following RS at Google DeepMind. Prev: Harvard, MIT, NASA, Google Brain.Mohammad Raihan Uddin @RaihanAkash0
271 Followers 4K Following Researcher- ML, AI, Federated Learning.Dustin Tran @dustinvtran
40K Followers 649 Following Research Scientist at Google DeepMind. I lead evaluation at Gemini / Bard. AI, Bayesian statistics, deep learning.citezenb | citezenb.t.. @CitezenB
15K Followers 16K Following Passionate about @Tezos, a blockchain of cultural significance. #blockchain #artchain https://t.co/c7t0jFOruQAmani @Sleausm176155
156 Followers 4K Following See the world on the road, and get to know yourself on the way!John @Teasto157063
495 Followers 5K Following See the world on the road, and get to know yourself on the way!Ayça Takmaz @aycatakmaz
328 Followers 391 Following PhD student @ETH Zurich, Student Researcher @GoogleAIAlexey Nekrasov @kumuji
261 Followers 525 Following Ph.D. student from @RWTH Love robotics, research and computer visionDaniel Marczak @danie1marczak
125 Followers 399 Following PhD Student @ Warsaw University of Technology && IDEAS NCBR | Continual Learning | Self-Supervised LearningPatrick Haller @padraiglindrome
264 Followers 940 Following PhD Student in Computational Linguistics @cl_uzh. Interested in language modeling, human language processing... and drag race I guess. he/himMichal Wolski @michalwols
647 Followers 1K Following Interested in large scale image recognition and retrieval Principal ML Eng at MyFitnessPal, prev @columbia @clarifai @nyufuturelabs @biteai (acquired by MFP)Angéline Pouget @angelinepouget
23 Followers 82 Following Student Researcher @ Google DeepMind | Data Science MSc @ ETH Zürich | 2022 Excellence ScholarBlingeria @Blingeriashop
37 Followers 277 Following High quality clothing and jewelry for a fair priceDerek @Toughth329832
463 Followers 5K Following See the world on the road, and get to know yourself on the way!Rong Ching Chang @AnnCC12
671 Followers 5K Following Fascinated by ML, LLMs, GNN, Multimodal models in social media. Ph.D. student @ucdavisHicham @chicham
75 Followers 2K FollowingOmerFaruk TAL @OmerFarukTal
6 Followers 144 FollowingEdgeAI Geek @edgeaiguy
1K Followers 5K Following Crafting AI solutions for tiny devices. | Ex-Samsung |Adam @adam31416
23 Followers 116 Following+ @tfius
962 Followers 4K Following Es may B my bae chord. Synthetic Cypherpunk Space Che Human. Decentralize it. Open Source Absolutist of Ontological Anarchy. transparent/invisible. Be.Vincent Lordier @vlordier
554 Followers 4K Followingrealmax @realmax17651231
1 Followers 57 FollowingAlpay Ariyak @AlpayAriyak
1K Followers 1K Following 𝗔𝗜 @RunPod_io | 𝗟𝗲𝗮𝗱: @OpenChatDev (𝟲𝟬𝟬𝗸+ 𝗱𝗼𝘄𝗻𝗹𝗼𝗮𝗱𝘀 on HuggingFace🤗)HinePo @Hine__Po
206 Followers 424 Following Head of AI & Data. Data science tech lead. Chemical engineer. Kaggle Competitions Expert (top 1%).AK @_akhaliq
307K Followers 3K Following AI research paper tweets, ML @Gradio (acq. by @HuggingFace 🤗) dm for promo follow on Hugging Face: https://t.co/q2Qoey80GxAndrej Karpathy @karpathy
974K Followers 904 Following 🧑🍳. Previously Director of AI @ Tesla, founding team @ OpenAI, CS231n/PhD @ Stanford. I like to train large deep neural nets 🧠🤖💥Lucas Beyer (bl16) @giffmana
56K Followers 443 Following Researcher (Google DeepMind/Brain in Zürich, ex-RWTH Aachen), Gamer, Hacker, Belgian. Mostly gave up trying mastodon as [email protected]Google DeepMind @GoogleDeepMind
941K Followers 275 Following We’re a team of scientists, engineers, ethicists and more, committed to solving intelligence, to advance science and benefit humanity.AI at Meta @AIatMeta
526K Followers 255 Following Together with the AI community, we are pushing the boundaries of what’s possible through open science to create a more connected world.Dmytro Mishkin 🇺�.. @ducha_aiki
18K Followers 590 Following Marrying classical CV and Deep Learning. I do things, which work, rather than being novel, but not working.Ross Wightman @wightmanr
18K Followers 1K Following Computer Vision @ 🤗. Ex head of Software, Firmware Engineering at a Canadian 🦄. Currently building ML, AI systems or investing in startups that do it better.Jeremy Howard @jeremyphoward
221K Followers 5K Following 🇦🇺 Co-founder: @AnswerDotAI & @FastDotAI ; Hon Professor: @UQSchoolITEE ; Digital Fellow: @StanfordSoumith Chintala @soumithchintala
185K Followers 871 Following Cofounded and lead @PyTorch at Meta. Also dabble in robotics at NYU. AI is delicious when it is accessible and open-source.Joan Puigcerver @joapuipe
864 Followers 375 Following Software Engineer in Research at Google DeepMind, Zürich.👩💻 Paige Bai.. @DynamicWebPaige
59K Followers 2K Following ✨Keep it simple, make it scale. AI should be about empowering people, building understanding, & making dreams realities. 👩💻GenAI @GoogleDeepMind ex-@GitHubrohan anil @_arohan_
12K Followers 2K Following Principal Engineer, @GoogleDeepMind Gemini. prev PaLM-2. Tinkering with optimization and distributed systems. opinions are my own.Basil Mustafa @_basilM
1K Followers 130 Following researching ML @ google brain ZRH | no strong opinions about AI, but very strong opinions about why herbal infusions are awful and should not be called teasHorace He @cHHillee
23K Followers 445 Following Working at the intersection of ML and Systems @ PyTorch "My learning style is Horace twitter threads" - @typedfemaleGoogle AI @GoogleAI
2.2M Followers 23 Following Google AI is focused on bringing the benefits of AI to everyone. In conducting and applying our research, we advance the state-of-the-art in many domains.Jürgen Schmidhuber @SchmidhuberAI
106K Followers 0 Following Invented principles of meta-learning (1987), GANs (1990), Transformers (1991), very deep learning (1991), etc. Our AI is used many billions of times every day.François Fleuret @francoisfleuret
30K Followers 477 Following Prof. @Unige_en, Adjunct Prof. @EPFL_en, Research Fellow @idiap_ch, co-founder @nc_shape. AI and machine learning since 1994. I like reality.Ferenc Huszár @fhuszar
40K Followers 1K Following Secular Bayesian. Associate Professor in Machine Learning @Cambridge_CL. Talent aficionado at https://t.co/RbJkoLguey Alum of @Twitter, Magic Pony and @BaldertonFabian Mentzer @mentzer_f
2K Followers 204 Following Senior Research Scientist at Google during the day, modular synth guy at night: https://t.co/Ea84aix9PmGuido van Rossum @gvanrossum
290K Followers 493 Following Python's BDFL-emeritus, Distinguished Engineer at Microsoft, Computer History Fellow, fully vaccinated. Opinions are my own. He/him.Shane Legg @ShaneLegg
51K Followers 57 Following Co-founder and Chief AGI Scientist, Google DeepMindJason Crawford @jasoncrawford
33K Followers 3K Following Founder, @rootsofprogress. I write about the history of technology and the philosophy of progress. Working to build the progress community and movementChris Olah @ch402
90K Followers 173 Following Reverse engineering neural networks at @AnthropicAI. DMs open! Previously @distillpub, OpenAI Clarity Team, Google Brain. Personal account.Rishabh Agarwal @agarwl_
6K Followers 539 Following Senior Research Scientist, @GoogleDeepMind, ex-🧠. Agents that make decisions. NeurIPS Best Paper (RLiable). Mila, IIT Bombay.Piotr Padlewski @PiotrPadlewski
1K Followers 319 Following Chief Meme Officer @ https://t.co/CtBrcKmliI, ex-Google Deepmind/Brain ZurichGagan Madan @_gaganm
273 Followers 562 Following Previously Research Eng @GoogleAI, Eng @ GPay India. Probably approximately incorrectMichael Levin @drmichaellevin
39K Followers 2K Following Scientist at Tufts University; my lab studies anatomical and behavioral decision-making at multiple scales of biological, artificial, and hybrid systems.Eyvind Niklasson @eyvindn
835 Followers 2K Following research @ google, working on self-organising systemsEttore Randazzo @RandazzoEttore
1K Followers 230 Following Researcher @ Google. I work on self-organising systems and artificial life.🦎Eliezer Yudkowsky ⏹.. @ESYudkowsky
175K Followers 89 Following The original AI alignment person. Missing punctuation at the end of a sentence means it's humor. If you're not sure, it's also very likely humor.Ben Poole @poolio
17K Followers 1K Following research scientist at google brain. phd in neural nonsense from stanford.Decoding The Gurus @GurusPod
11K Followers 75 Following Decoding the Gurus Podcast with @ArthurCDent and @C_Kavanagh. Email at: [email protected]. Support the pod at: https://t.co/9I87qa9sT1Michael Tschannen @mtschannen
1K Followers 616 Following Machine learning researcher @GoogleDeepMind. Past: @Apple, @awscloud AI, @ETH_en. Multimodal/representation learning.Mathilde Caron @mcaron31
1K Followers 27 Following Research Scientist @googIeresearch Grenoble ⛰️ Previously PhD student @Inria & @MetaAI (FAIR)Jacob Buckman @jacobmbuckman
5K Followers 373 Following Founder @manifest__ai. PhD candidate @MILAMontreal. Formerly @jhuclsp, @GoogleAI, @SCSatCMU.Durk Kingma @dpkingma
35K Followers 346 Following Deep learning, mostly generative models. Prev. Google Brain/DeepMind, founding team @OpenAI. Inventor of the VAE, Adam optimizer, among other things. ML PhD.Cristian Garcia @cgarciae88
6K Followers 1K Following JAX/Flax at Google DeepMind | Open Source | 🇨🇴Internal Tech Emails @TechEmails
523K Followers 901 Following Internal tech industry emails that surface in public records. 🔍Max Woolf @minimaxir
19K Followers 459 Following Data Scientist at @BuzzFeed in San Francisco // AI content generation R&D // Mastodon: @[email protected]Jordan Schneider @jordanschnyc
47K Followers 4K Following Newsletter: https://t.co/j23uuLxE59 join 30k readers Podcast: https://t.co/4sx8iev5Az Business partnerships: https://t.co/a8hYrI8Qde Fellow: @cnasdcJoelle Pineau @jpineau1
10K Followers 348 Following AI researcher. VP AI Research (FAIR), @AIatMeta. Professor of Computer Science, @mcgillu. Core academic member, @Mila_QuebecXiao Wang @brainshawn
130 Followers 25 Following Reseacher in @GoogleDeepMind Zurich current: vision-language & data-centric research; 2015-2020: text understanding; before 2015: distributed systemsDavid Duvenaud @DavidDuvenaud
27K Followers 3K Following Machine learning prof @UofT. Working on generative models, inference, & latent structure.Markus Kneer @kneer
1K Followers 2K Following Professor of Ethics of Artificial Intelligence. I study cognition, language and norms (both in human/human and human/AI interaction).Karsten Kreis @karsten_kreis
2K Followers 443 Following Senior Research Scientist at @NVIDIA | Former Physicist | Deep Generative Learning. Opinions are my own.Giada Pistilli @GiadaPistilli
9K Followers 584 Following Principal Ethicist @HuggingFace | Philosophy Ph.D. @Sorbonne_Univ_ & @CNRSHugo Larochelle @hugo_larochelle
113K Followers 625 Following Google DeepMind researcher, machine learning professor, ex-Twitter Cortex, father of 4, wine/music/comedy enthusiastJonathan Ho @hojonathanho
4K Followers 151 FollowingChitwan Saharia @Chitwan_Saharia
3K Followers 289 Following @ideogram_ai Past: Sr. Research Scientist @GoogleAI || B. Tech, CSE, @IITBombayAccepted papers at TM.. @TmlrPub
3K Followers 2 Followinggandamu @gandamu_ml
16K Followers 5K Following Prev https://t.co/a96FYiLT41 · https://t.co/ren7Ov9vxx. Music videos: https://t.co/iFubkxDg5gMIT CSAIL @MIT_CSAIL
297K Followers 22K Following MIT's Computer Science & Artificial Intelligence Laboratory (CSAIL). Media Inquiries: [email protected]Raphaël Millière @raphaelmilliere
10K Followers 2K Following Philosopher of Artificial Intelligence & Cog Science @Macquarie_Uni Past @Columbia @UniofOxford Also on other platforms Blog: https://t.co/2hJjfSid4ZMichelle R Carney @michellercarney
3K Followers 3K Following UXR for ML @TensorFlow @Google. Founder @mluxmeetup. Lecturer @stanforddschool Member @feminist_ai. Former Fellow @AFOGBerkeley. All views my own. she/herTensorFlow.js @TensorFlow__JS
137 Followers 1 Following TensorFlow.js brings Machine Learning to JavaScript. What will you make? This account is a temp work account, follow @jason_mayes for latest news on #WebMLJeremy Torman (mintin.. @TormanJeremy
26K Followers 11K Following Ai art since 2014 | Featured at TED's 2023 conference | Sold at Sotheby's |Excited to share our #ICLR2024 paper, focused on reducing bias in CLIP models. We study the impact of data balancing and come up with some recommendations for how to apply it effectively. Surprising insights included! Here are 3 main takeaways.
It is an amazing time to work in the cognitive science of language. Here are a few remarkable recent results, many of which highlight ways in which the critiques of LLMs (especially from generative linguistics!) have totally fallen to pieces.
@_akhaliq I added OWL-ViT v2 to the plot. A single OWLv2 B/16 model, finetuned on O365+VG, covers all speed/accuracy combinations: Simply adjust the inference resolution to match your latency requirements. No re-training needed. arxiv.org/abs/2306.09683
📢@AndreasPSteiner and I will present Image Captioners Are Scalable Vision Learners Too @NeurIPSConf today! Talk: 3:35pm, R02-R05 (level 2) Poster: 5pm, #214 💾We also just released code as part of big_vision: github.com/google-researc… 1/2
Whenever I have issues w/ training big models for CV (but not only) tasks, most of the times I find my answer or a part of it in a corner or plot of a paper of this "(sub)team" producing as much scientific knowledge as a pile of teams. Kudos @giffmana @XiaohuaZhai @__kolesnikov__
Here's what our (sub)team in Zürich has done for OSS vision over the past 5y, besides inventing ViT: 1) Make i21k a thing Release: 2) best CLIP (siglip) by a large margin 3) best i1k ResNet50 ever 4) best pre-trained ResNets 5) >55k ViTs 6) Most efficient JAX/TPU CV code deets👇
Here's what our (sub)team in Zürich has done for OSS vision over the past 5y, besides inventing ViT: 1) Make i21k a thing Release: 2) best CLIP (siglip) by a large margin 3) best i1k ResNet50 ever 4) best pre-trained ResNets 5) >55k ViTs 6) Most efficient JAX/TPU CV code deets👇
Introducing Soft MoE! Sparse MoEs are a popular method for increasing the model size without increasing its cost, but they come with several issues. Soft MoEs avoid them and significantly outperform ViT and different Sparse MoEs on image classification. arxiv.org/abs/2308.00951
Wow! What a substrate for running NCAs!
Ever wanted to easily cover objects with RGB LEDs? Now you can with PCBend, our latest project accepted at @siggraph with @Manas161997, C. Schreck, @HugronPa, @bernd_bickel, @sylefeb. Stay tuned for more over the coming weeks! #SIGGRAPH2023 #neopixel #PCB 1/3⬇️
@giffmana @MarkPKCollier @RJenatton @_basilM @brainshawn @XiaohuaZhai @AndreasPSteiner @jesse_berent @GoogleAI No problem at all! Mark and I will present a poster at the workshop in person though :)
Finally ready to share Biomaker CA, a Biome Maker project using Neural Cellular Automata. w/ @zzznah See live article google-research.github.io/self-organisin… (with lots of videos) and arxiv paper arxiv.org/abs/2307.09320. 1/N
🥈at Ironman Switzerland! Overwhelmed with the pace of AI development? An engaging hobby is a great way to stay enthusiastic. For me, it's endurance training. Bonus: long rides are a perfect time to ruminate on research ideas!
Check out NaViT, a native resolution ViT for all aspect ratios, enhancing training efficiency & performance. By preserving aspect ratios, it improves fairness-signal annotation, useful where metrics like group calibration are noise-sensitive. NaViT helps overcome such challenges.
Patch n' Pack: NaViT, a Vision Transformer for any Aspect Ratio and Resolution paper page: huggingface.co/papers/2307.06… The ubiquitous and demonstrably suboptimal choice of resizing images to a fixed resolution before processing them with computer vision models has not yet been…
Do you want to accelerate your vision model without losing quality? NaViT takes images of arbitrary resolutions and aspect ratios - no more resizing to square with constant resolution. One cool implication is that you can control compute/quality tradeoff by resizing:
Patch n' Pack: NaViT, a Vision Transformer for any Aspect Ratio and Resolution paper page: huggingface.co/papers/2307.06… The ubiquitous and demonstrably suboptimal choice of resizing images to a fixed resolution before processing them with computer vision models has not yet been…
Other method of reducing compute is dropping random tokens. This turned out to be much worse strategy, as the quality drops much faster
Really excited to share NaViT - our latest work on *efficiently* extending ViTs to handle variable resolution and variable aspect ratio images. We're bringing back one of the cool features of ResNets - the ability to apply them to different image sizes at train at test time.
Patch n' Pack: NaViT, a Vision Transformer for any Aspect Ratio and Resolution paper page: huggingface.co/papers/2307.06… The ubiquitous and demonstrably suboptimal choice of resizing images to a fixed resolution before processing them with computer vision models has not yet been…
@neilhoulsby @m__dehghani @_basilM @JonathanHeek @MJLM3 @mcaron31 @AndreasPSteiner @joapuipe @ibomohsin @avitaloliver @PiotrPadlewski I really think you're underselling the third chart. The x axis is logarithmic, these are very serious efficiency gains for applications. Amazing job
You’re training a computer vision model. Want improved efficiency? Or maybe arbitrary aspect ratios? Perhaps process any resolution? …why not all the above? We’re thrilled to present NaViT, so you can have your cake & eat it (at any size or aspect ratio) arxiv.org/abs/2307.06304