Sanbu GenAI (@sanbuphy) 's Twitter Profile
Sanbu GenAI

@sanbuphy

Image GenAI / LLM / Open Source Community
Lower the threshold for people to attain happiness.

ID: 1611582713550770177

linkhttps://github.com/sanbuphy calendar_today07-01-2023 04:37:24

58 Tweet

13 Takipçi

249 Takip Edilen

Sanbu GenAI (@sanbuphy) 's Twitter Profile Photo

FramePack is here! ✨ Official implementation for "Packing Input Frame Context" - a next-frame prediction structure for video gen! 🔥 Compresses context, runs efficiently on 6GB GPU (even 13B models!) #VideoAI #AIresearch #VideoGeneration #FramePack

FramePack is here! ✨ Official implementation for "Packing Input Frame Context" - a next-frame prediction structure for video gen! 🔥 Compresses context, runs efficiently on 6GB GPU (even 13B models!)  #VideoAI #AIresearch #VideoGeneration #FramePack
Sanbu GenAI (@sanbuphy) 's Twitter Profile Photo

The precise parameter modulation in mmdit blocks is truly fascinating! Is it necessary to have such fine-grained control for generating good results? Doesn't this go against the idea of simple big data? What's the necessity of this complexity?

The precise parameter modulation in mmdit blocks is truly fascinating! Is it necessary to have such fine-grained control for generating good results? Doesn't this go against the idea of simple big data? What's the necessity of this complexity?
Sanbu GenAI (@sanbuphy) 's Twitter Profile Photo

Alibaba's Qwen3 is about to drop! 🚀 Expect big things from Qwen3-8B and Qwen3-MoE-15B-A2B, including potential multi-modal breakthroughs. Can't wait to see the competition heat up with this release! Alibaba Cloud is making huge strides in open-source AI. #Qwen3 #AI #ML #Tech

Alibaba's Qwen3 is about to drop! 🚀 Expect big things from Qwen3-8B and Qwen3-MoE-15B-A2B, including potential multi-modal breakthroughs. Can't wait to see the competition heat up with this release! Alibaba Cloud is making huge strides in open-source AI. #Qwen3 #AI #ML #Tech
Sanbu GenAI (@sanbuphy) 's Twitter Profile Photo

This open-source LLM for Lean 4 theorem proving is next level. Recursive theorem proving pipeline powered by DeepSeek-V3. This is going to be a game-changer for formal theorem proving in AI. Github link: github.com/deepseek-ai/De…

This open-source LLM for Lean 4 theorem proving is next level. Recursive theorem proving pipeline powered by DeepSeek-V3. This is going to be a game-changer for formal theorem proving in AI. Github link: github.com/deepseek-ai/De…
Sanbu GenAI (@sanbuphy) 's Twitter Profile Photo

OBREAKING: Gemini Pro 2.5 becomes FIRST AI Pokémon League Champion. New benchmark be like: How fast can my LLM finish this game ? 📊 Pokemon token efficiency > LLM bench now. Live reaction here ➡️ twitch.tv/gemini_plays_p… #AIgaming #GeminiPro #Pokemon #ArtificialIntelligence

OBREAKING: Gemini Pro 2.5 becomes FIRST AI Pokémon League Champion.  New benchmark be like: How fast can my LLM finish this game ? 📊 Pokemon token efficiency > LLM bench now.  Live reaction here ➡️ twitch.tv/gemini_plays_p… #AIgaming #GeminiPro  #Pokemon #ArtificialIntelligence
Sanbu GenAI (@sanbuphy) 's Twitter Profile Photo

🔥 Stanford CS336 : Language Modeling from Scratch,the ultimate bootcamp for building LLMs from ground up. Implementing a Transformer? FlashAttention in Triton? Real talk on DPO/GPO and handling Common Crawl? No more magic black-box LLMs. #Stanford #LLM #AI

🔥 Stanford CS336 : Language Modeling from Scratch,the ultimate bootcamp for building LLMs from ground up. Implementing a Transformer? FlashAttention in Triton? Real talk on DPO/GPO and handling Common Crawl?   No more magic black-box LLMs. #Stanford #LLM #AI
Yi Xu (@_yixu) 's Twitter Profile Photo

🚀Let’s Think Only with Images. No language and No verbal thought.🤔 Let’s think through a sequence of images💭, like how humans picture steps in their minds🎨. We propose Visual Planning, a novel reasoning paradigm that enables models to reason purely through images.

🚀Let’s Think Only with Images.

No language and No verbal thought.🤔 

Let’s think through a sequence of images💭, like how humans picture steps in their minds🎨. 

We propose Visual Planning, a novel reasoning paradigm that enables models to reason purely through images.
Chrome for Developers (@chromiumdev) 's Twitter Profile Photo

Announcing our public preview of Chrome DevTools MCP! Experience the full power of DevTools in your AI coding agent → goo.gle/4pDE6Tk With Chrome DevTools MCP, your AI agent can run performance traces, inspect the DOM, & perform real-time debugging of your web pages.

Sanbu GenAI (@sanbuphy) 's Twitter Profile Photo

NotebookLM :Tiny Google Labs team → under 2 months → idea to viral hit. Key takeaways for AI product crews: Ship fast, iterate with users. AI-native UX, not AI bolt-on. Structure-first (NotebookLM’s 3-panel flow = read→chat→create). Context-aware, tools appear when needed.

NotebookLM :Tiny Google Labs team → under 2 months → idea to viral hit.
Key takeaways for AI product crews:
Ship fast, iterate with users.
AI-native UX, not AI bolt-on.
Structure-first (NotebookLM’s 3-panel flow = read→chat→create).
Context-aware, tools appear when needed.