bidhan ✈️ CVPR

ceo @bageldotcom. previously code monkey at amazon, cashapp, instacart. hiring - dms are open.

[bidhan ✈️ CVPR @bidhan]

We're releasing Paris 2.0, which, to our knowledge, is the world's first decentralized trained video generation model.

We benchmarked it against a monolithic model trained on the same data and compute budget, and Paris 2.0 outperformed the monolithic by ~2x on FVD benchmark.

[bidhan ✈️ CVPR @bidhan]

@jasonnov @CVPR 🫡

[bidhan ✈️ CVPR @bidhan]

heading to @CVPR with the Bagel Labs team. if you want some elite merch (we spent a month designing it), discuss world models, physical ai, distributed training -- I'm your guy 🫡

[bidhan ✈️ CVPR @bidhan]

heading to @CVPR with the Bagel Labs team. if you want some elite merch (we spent a month designing it), discuss world models, physical ai, distributed training -- I'm your guy 🫡

[Ankur Nagpal @ankurnagpal]

If you’re over the age of 30 and going to tech week parties

Why

[bidhan ✈️ CVPR @bidhan]

@natolambert @TheZachMueller well said 🫡

[bidhan ✈️ CVPR @bidhan]

what if I told you there’s a subsection of the AI industry which has 1000x the TAM of LLMs?

[bidhan ✈️ CVPR @bidhan]

a friend let me know that bagel labs was trending yesterday

[bidhan ✈️ CVPR @bidhan]

We're releasing Paris 2.0, which, to our knowledge, is the world's first decentralized trained video generation model.

We benchmarked it against a monolithic model trained on the same data and compute budget, and Paris 2.0 outperformed the monolithic by ~2x on FVD benchmark.

[bidhan ✈️ CVPR @bidhan]

@yacinelearning man, we were in the TRENCHES

[bidhan ✈️ CVPR @bidhan]

@yacinelearning thank you my friend 🙏

[bidhan ✈️ CVPR @bidhan]

We're releasing Paris 2.0, which, to our knowledge, is the world's first decentralized trained video generation model.

We benchmarked it against a monolithic model trained on the same data and compute budget, and Paris 2.0 outperformed the monolithic by ~2x on FVD benchmark.

[bidhan ✈️ CVPR @bidhan]

@Camcamera11 indeed. 2x better against a monolithic model.

[bidhan ✈️ CVPR @bidhan]

@jasonnov good question! we explored that here. tldr is monolithic diffusion models have conflicting data for any given input. when you distribute them into "experts" the conflict between input data and experts become less. https://t.co/hdj9YWcBrS

[bidhan ✈️ CVPR @bidhan]

@TziokasV thanks Vassilis!

[bidhan ✈️ CVPR @bidhan]

@MTSlive missed the @bageldotcom launch in recap 🫡

https://t.co/i2AisKHxqt

[bidhan ✈️ CVPR @bidhan]

We're releasing Paris 2.0, which, to our knowledge, is the world's first decentralized trained video generation model.

We benchmarked it against a monolithic model trained on the same data and compute budget, and Paris 2.0 outperformed the monolithic by ~2x on FVD benchmark.

[bidhan ✈️ CVPR @bidhan]

@gregosuri @bageldotcom thanks for sharing Greg!

[bidhan ✈️ CVPR @bidhan]

@supremebeme wait, that's a good idea for the next model

[bidhan ✈️ CVPR @bidhan]

@svpino Thanks for the kind words Santiago!

[bidhan ✈️ CVPR @bidhan]

@TomLikesRobots yes, it is indeed done for the first time by @bageldotcom team

[bidhan ✈️ CVPR @bidhan]

come to the bagel labs paper session and booth at CVPR!

[Gin Jiang @CVPR @ZhiyingJ]

we found decentralized diffusion framework work for video generative models too! Still an early attempt, and a lot of open research questions left to explore. Would love to dig into it more next week at CVPR :)

[bidhan ✈️ CVPR @bidhan]

@jbrukh @bageldotcom Thanks Jake! 🫡🥯

[bidhan ✈️ CVPR @bidhan]

@Omarboucher that's the spirit 🫡