Agajan Torayev

@mervenoyann this is true. you will hear a lot "my love", "my darling", etc. so polite.

people of London are insanely nice

I think @karpathy means cursor for Beamer (LaTeX)

@mervenoyann @huggingface different versions of packages, lack of backward compatibility

hey folks 👋🏻 what are the worst friction points in @huggingface ecosystem (transformers, Hub etc) when you're building vision/multimodal stuff? I'd love to hear and see if we can help 🙏🏻

@morazow @MetaforDevs @cerebral_valley Minnetdar 😊

@torayeff @MetaforDevs @cerebral_valley Nice, congratulations 🥳

Last week, I was in SF and, by a lucky chance, participated in the first LlamaCon Hackathon — and surprise: I won 3rd place and a $6,000 cash prize among top devs in Silicon Valley 🎉 Huge thanks to @MetaforDevs and @cerebral_valley for hosting this great event!

We're excited to announce the winners of the first LlamaCon Hackathon! These talented individuals and teams have demonstrated exceptional skill and creativity in their projects using Llama. 🥇 1st Prize: OrgLens An AI-powered expert matching system that connects you with the right professionals within your organization. By leveraging data from various sources, OrgLens creates a comprehensive knowledge graph and detailed profiles, streamlining expert matching. See their GitHub Repository: bit.ly/4k62WaO @KPJedrzejewski, @clemhus 🥈 2nd Prize: Compliance Wizards An AI-powered transaction analyzer designed to detect fraud and alert users. It uses Llama API’s multi-modality to assist fraud assessors in determining client involvement in criminal activities. See their GitHub Repository: bit.ly/3RRxiBS @SamDc73, @k_a__reem, @nicetomeetyu2, @sorhanft 🥉 3rd Prize: Llama CCTV Operator A Llama CCTV AI control room operator that identifies custom surveillance video events without model fine-tuning. It uses Llama 4’s multi-modal image understanding to assess and report predefined events. See their GitHub Repository: bit.ly/4d9UPrw @torayeff 🌟 Best Llama API Usage: Geo-ML This project uses Llama 4 Maverick and GemPy to generate 3D geological models, processing extensive geology reports into structured data for 3D representations. See their GitHub Repository: bit.ly/3GITT15 @WilliamJSDavis Please join us in congratulating these winners on their outstanding achievements! We're honored to have them as part of the Llama community. 🎉

Hacking with the best at #LlamaconHackathon25 @metafordevelopers

sometimes i wonder, what is the point of it all. o3 is killing it on humanity's last exam. then i try to use the llm to automate my job and i remember how useless benchmarks are

"scrum master" in people's profiles somehow reads like "scum master"

@orr_zohar @_akhaliq Hey @orr_zohar why the weights are gone? I remember you had github repo also, cannot find it anymore.

@_akhaliq Thank you @_akhaliq for highlighting our work! We hope that Apollo's 🧑‍🚀insights help further accelerate the video-LMM field with better, more informed design decisions! 🚀

Meta releases Apollo An Exploration of Video Understanding in Large Multimodal Models a family of state-of-the-art video-LMMs

@ARChaser_ The demo looks really cool! Are you planning to release the code anytime soon?

The demo has been resumed. 😉

🎉 Online Demo Since TAPTRv3 is an online tracker, recently we have implemented a streaming inference mode, which allows us to process videos of any length on RTX3090! With the support of @Gradio and @huggingface, we have deployed the demo at huggingface.co/spaces/HYeungL…. Try it out!

@burkov @StackOverflow @StackOverflow was king. I was once banned from SO, but I loved it because I learned from the best in the field. Sorry to hear that you had a different experience.

No website decline leaves me feeling as satisfied as @StackOverflow. The people in this place detest you so much, they don't even try to hide it. They will reject your questions, prevent you from answering, and openly mock you in the comments. The toxicity level of these people surpasses any reasonable standard. Burn this place to the ground.

@iruletheworldmo decode 18735

someone from ssi just followed me. intriguing is all.

@EurekaLabsAI @BCiechanowski has very good visually enhanced tutorials/explanations

Curious what is the content (tutorial, book, videos,...) that sparked the most Eurekas for you recently and why?

@erenbali 🙋‍♂️

Raise your hand 🙋‍♂️ if you dragged young kids thousands of miles to a city with the most stunning beaches of the worlds just to remember they like the pool more

@natfriedman torayeff.com inspired by you :)

A little torn that my greatest contribution to tech might be bringing back 1997-style times new roman websites.

@fofrAI Impressive! It generates images of "identical twins"— "similarly looking" but not "exactly the same person." Differences are noticeable with very familiar faces (family members, for example), but for strangers, the generated images appear consistent and like the same person.

Model: replicate.com/fofr/consisten… Code: github.com/fofr/cog-consi… Each run of the model loops through different poses, completing the same workflow many times (rather than duplicating within the workflow)

Here's the first release of my consistent-character model on Replicate. It uses InstantID, IPAdapter, Controlnet and FaceDetailer, with SDXL Lightning. Prompt clothes and hairstyle for best consistency It's open source but non-commercial, all links below 👇

@thatstraw Mandriva

Did everyone start their Linux journey with Ubuntu? 🤔

@iScienceLuvr Here again the GitHub link is broken github.com/AlaaLab/Dr-LLa…

Dr-LLaVA: Visual Instruction Tuning with Symbolic Clinical Grounding abs: arxiv.org/abs/2405.19567 Construct synthetic multi-turn multimodal conversational datasets based on symbolic rules written by clinicians, diversity with GPT-4 paraphrasing. Train LLaVA model with RLHF using a reward function checking clinical correctness and consistency with the symbolic rules. Example of approach provided for analyzing bone marrow pathology slides.

If you haven't noticed, even if a paper has a link to a repo, I won't share it in a paper tweet if there actually isn't any code in it. Too many times researchers create placeholder repos with unfulfilled promises of releasing code. This needs to stop!