
Research @stanford & @lmu_muenchen. Prev Kaza, Partex. Occasionally I write at https://nitter.cf/t.co/Yle2ATucrc
Munich & SF
Joined June 2024
- Tweets579
- Following239
- Followers135
- Likes731
Great meeting @nikhilkamathcio and @thetanmay
Thank you for being unscripted and honest haha :)
3
155
Very good talk @nikogrupen on how does it how environment design at Harvey.
Synthetic data it is - but yes it needs a prior / anchor for the same.
2
98
Krrish Agarwalla retweeted
Over the past week, I’ve been overwhelmed by the love and support of friends, industry colleagues, and people I've never met. Thank you to everyone who has shown kindness to my family and me. It means the world to me.
I came to the United States 53 years ago, and becoming an American citizen is one of the greatest gifts and proudest moments of my life. NVIDIA has been my life’s work. Building it into the company it is today alongside my colleagues has been a privilege and an honor. Nowhere but in America could a company like NVIDIA be possible. I love our country, and I love NVIDIA.
I did not live up to the standards expected of everyone at NVIDIA, and I have made serious mistakes in my personal life. I created problems for the Board that I deeply regret. I own my mistakes completely, and I am committed to cooperating with the Board’s independent investigation.
Since my departure, I’ve seen an outpouring of calls from NVIDIA’s employees, investors, business partners, and customers for me to return as CEO. Many continue to reach out to me directly as the company navigates complex technology, market, competitive, and geopolitical dynamics. NVIDIA is a once-in-a-generation company. If the Board of Directors decides that my return is in the best interest of our shareholders, our employees, the company, and America, I would be honored to answer these calls and return as CEO.
Thank you again to everyone who has shared their good wishes. I am profoundly humbled by your support. To my family: I love you, and I will spend the rest of my life being worthy of yours.
Jensen
I think these days this is the cost to get a paper into a good conference.
Any opinion folks ?
1
6
1,009
I really enjoyed the hackathon as well. There was an insane amount of crazy ideas flowing through.
We built an autonomous loop where ripple could quantize itself when paired up with open evolve framework.
We got the size of model from ~60GB ( NVFP4 ) to 30 GBs of model size without compromising much on model quality.
Some details here :
nitter.cf/Krrishagar/status/2122…
The first @RicursiveAI hackathon, centered on Ripple-65B Preview, was so much fun! Participants built impressive computer use demos, a lightning-fast IDE built and operated by Ripple, kernel designs that outperformed Luna, and more.
We never directly optimized the model for some of these capabilities, like some cool visual demos.
One of the most important things in development of local coding agents is that we should be able to serve the modal locally -> there are many ways to get here but one of the most important techniques in serving is quantization.
At the @RicursiveAI hackathon, I wanted to understand that whether we can let an agent loose to quantize itself i.e. can ripple 65B quantize itself.
As fundamentally quantization is a search problem, you search across which layers to quantize and which layers to leave out.
Thread 1/n 🧵
Limitations :
- Severely limited by compute
- Better design of the optimization objective to get to our goal better of lower bits but higher accuracy because at the end it started to push towards lower bits and compromised alignment with the original model.
- I have not tested it to AIME and other benchmarks this, being how good KL is a proxy will be tested through it.
Thank you @talaei_shayan @ethanboneh @TutramN for organizing such a fun hackathon. I learned a lot in general and about @RicursiveAI as well.
I did a good trip around Golden gate park today :)
It was fun, let's call it gg park from now on.
3
172
Doing a hackathon at @RicursiveAI -> trying to build an agent which can quantize any model helping it run faster while being very accurate :)
This is one of the ways agents can be used for discovery :0
4
131














