Danilo Danese

3 papers in 5 days. Authors shall cite each other and consolidate all findings :) #recsys

Thrilled to announce our paper, "Do Recommender Systems Really Leverage Multimodal Content?", has been accepted at #CIKM2025! 🔥 We investigate if the gains in Multimodal RecSys come from true understanding or just bigger models. A deep dive into what really works. (Thread 👇)

I want to extend a special and heartfelt thank you to my main co-authors, @MattAttimonelli and @danilo_danese. This paper is a direct result of their brilliant insights and deep commitment. I am incredibly proud of what we have accomplished together! 🙌

@cyrilzakka @argmax @awnihannun Everything works well! Nice Job!

@argmax 5) On-device LLM generation now powered by MLX and available with a quick keyboard shortcut. Let us know what you think! /cc @awnihannun

Christmas came early! 🎅🏻 Today marks the newest release of the HuggingChat 🤗 update with some really exciting capabilities! First up, automatic context injection! 1) Open a file in a supported app, summon HFChat, and it pre-populates the context window. No more copy-pasting

🎉 Excited to announce our @IEEEAccess article! NeuroSense: EEG dataset for emotion recognition with low-cost devices. 🧠Available on GitHub 👉 github.com/sisinflab/Neur… 📚 Paper 👉 ieeexplore.ieee.org/document/10737… #EEG #EmotionRecognition #AffectiveComputing

@scne just presented their latest work at @ACMRecSys #GraphLearning session. This work, co-authored by @dmalitesta @alberto_mancino @walteranelli @TommasoDiNoia Explores the relationship between topological datasets characteristics and GNN based recommender systems.

The @ACMRecSys Session about Graph Learning has just started! Hosted by the one and only @sciueferrara

Day 3 of @ACMRecSys #recsys2024 #bari First session: Multi-task Learning

It's really happening! @ACMRecSys #recsys2024 #bari

Today @jnvinagre keynote presentation at @FAccTRec workshop as part of @ACMRecSys. The talk explores how EU regulations #DSA and #AIAct will shape the future of #RecSys research, highlighting the opportunities they bring. We look forward to seeing you there!

The #RecSysChallenge workshop has officially begun! Join us in the room “Aula Magna” for an exciting keynote by @Lindskow. Don't miss it! @ACMRecSys #recsys2024

From the entrance test for computer science engineering to organizing #RecSys2024, it feels like a blink of an eye 😂

@dome_271 Nice work, very clear explanation!

I finally released my new video on YouTube about Diffusion Models / Score-Based Generative Models. youtube.com/watch?v=B4oHJp… Literally planned this for a year and put so much work in. I think this approach to diffusion models is so intuitive and highly recommend giving that a go! Video is 38min long, so you will need some time to watch that haha.

Tomorrow I'm heading to Bari to give a lecture on #Multimodal #DeepLearning for #Recommendation at the #RecSys2024 summer school ✈️ 😎 📅 Save the date: Oct 09, 2024 📚 Material is already available online: github.com/danielemalites… See you very soon!

Open-MAGVIT2 An Open-Source Project Toward Democratizing Auto-regressive Visual Generation paper page: huggingface.co/papers/2409.04… We present Open-MAGVIT2, a family of auto-regressive image generation models ranging from 300M to 1.5B. The Open-MAGVIT2 project produces an open-source replication of Google's MAGVIT-v2 tokenizer, a tokenizer with a super-large codebook (i.e., 2^{18} codes), and achieves the state-of-the-art reconstruction performance (1.17 rFID) on ImageNet 256 times 256. Furthermore, we explore its application in plain auto-regressive models and validate scalability properties. To assist auto-regressive models in predicting with a super-large vocabulary, we factorize it into two sub-vocabulary of different sizes by asymmetric token factorization, and further introduce "next sub-token prediction" to enhance sub-token interaction for better generation quality. We release all models and codes to foster innovation and creativity in the field of auto-regressive visual generation.

@0xmaddie_ @multimodalart The dev model needs something like 40GB for 1024x1024 images

@multimodalart does this UI mention how much vram it uses? it looks like it says A100 so I guess flux uses like 24GB in 16bit, or can you quantize it down some more or what?

FLUX.1 [dev] space just dropped 🪩 huggingface.co/spaces/black-f…

Last Week, I presented at the World Conference on eXplainable Artificial Intelligence #xAI-2024. It has been very interesting to learn insights on xAI and meet other professionals in the field. 🧠🤖

The poster session at the #GenAI summer school (gemss.ai) is in progress and fantastic interactive! #GeMSS24

We have kicked off! The best summer school on #GenerativeAI: ✨Generative Modeling Summer School 2024✨! We turn @TUeindhoven into the world's capitol of #GenAI. Amazing lineup: @jesfrellsen, @pamattei, @dpkingma, @liyzhen2, @helibenhamu, @StephanMandt, @ewa_szczurek, @vlamen 🤩

@scne from @PolibaOfficial just presented the version **2.0** for our framework #Ducho at @TheWebConf 😎 Check out the code 💻(github.com/sisinflab/Ducho) and paper 📖 (dl.acm.org/doi/10.1145/35…) w/ @scne @MattAttimonelli @danilo_danese @TommasoDiNoia