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.
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 🫡
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 🫡
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.
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.
@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
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.
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 :)