Adarsh

@himanshustwts @smallest_AI Congratulations!

Career update: Excited to share that I have joined the incredible team at @smallest_AI to work on Research x Devrel! The team is cooking incredible small + efficient multi-modal models and it feels like an exciting time to push the frontier on scale!

Many things that seemed like eternities away feel possible with the advent of things like @claudeai! For example, I never had the courage to do any kernel work, let alone look into its source code. But I can now at least take specific help from Claude and be effective even when studying things. Of course, you cannot start from an empty slate here -- you still need some basic knowledge about computer architecture, processors, GPU programming model, etc. But even that can be structured these days. I am still learning to best use these powerful tools and I am absolutely enjoying the process.

This is going to be fire 🔥🔥

6 months ago, @taiuti and I made a bet. World models are the next platform shift, and we saw it coming. Since then, we've assembled a team from Apple, Meta, Google, Adobe & Microsoft and raised from top-tier investors. Come build with us: reactor.inc

The era of world models is here.

@taiuti @_bschmidtchen Congratulations guys 🥳🥳

Big update: I’m starting a new company. 6 months ago, @_bschmidtchen and I made a bet. What if entire worlds could be generated on the fly, pixel by pixel? World models are the next platform shift, and we saw it coming. Since then, we’ve: - secured major contracts across media and physical AI industries - assembled a team of 10+ from Apple, Meta, Google, Adobe & Microsoft - raised from top-tier investors More details soon. We’re scaling fast, and hiring now. Come build with us: reactor.inc

@ersascape Yeah I'm on jio. Cc @JioCare

@adarshxs seems to work here - could just be Jio

is github down or something?

Param-2 is live on @huggingface. 17B parameter MoE. 22 trillion tokens. 22 Indian languages. Reasoning, math, code, tool calling. Built from scratch, in India, open for the world. 🤗 Try it now: huggingface.co/bharatgenai/Pa… #BharatGen #Param2 #OpenSource

🪣 We just shipped Storage Buckets: S3-like mutable storage, cheaper & faster Git falls short for everything on high-throughput side of AI (checkpoints, processed data, agent traces, logs etc) Buckets fixes that: fast writes, overwrites, directory sync 💨 All powered by Xet dedup so successive checkpoints skip the bytes that already exist ➡️

The World Models era is here. AI that understands the world, and lets you create and interact with new ones, all in real time. More soon.

Introducing Modular Diffusers 🔥 The `DiffusionPipeline` abstraction in Diffusers has established a standard in the community. But it has also limited flexibility. Modular Diffusers breaks those shackles & enables the next gen. of creative user workflows 🧨 Details ⬇️

Congratulations @SarvamAI 🥳 Day-0 support on SGLang(@lmsysorg) Support PR: github.com/sgl-project/sg…

📢 Open-sourcing the Sarvam 30B and 105B models! Trained from scratch with all data, model research and inference optimisation done in-house, these models punch above their weight in most global benchmarks plus excel in Indian languages. Get the weights at Hugging Face and AIKosh. Thanks to the good folks at SGLang for day 0 support, vLLM support coming soon. Links, benchmark scores, examples, and more in our blog - sarvam.ai/blogs/sarvam-3…

There is no best VLM OCR model - rankings can flip completely by document type. I built ocr-bench: run open OCR models on YOUR documents, get a per-collection leaderboard. VLM-as-judge with Bradley-Terry ELO, all running on @huggingface. No local GPU needed.

the generated pixels are coming for your UI frameworks and game engines folks

[Hugging Face Machine Learning Club India] We are very excited to host @_lewtun as the speaker of our very first event. He has very kindly accepted to talk about QED-Nano. The talk is on 10th March 2000 Hrs IST. (Join the google meet, it will not be streamed) More details in the thread 🧵