For more details, check out our blogpost: https://t.co/PmU54rNoen, and our paper:

Powered by deterministic execution through our Reproducible Operators (RepOps) library, Verde enables verification of training, fine-tuning, and inference tasks across diverse hardware.

Today I had the opportunity to talk about verifying outsourced ML execution at the University of Twente. I discussed the landscape of verification methods and introduced our approach, Verde, at @gensynai. Thanks @Rymoah for the invitation!

1/ Introducing Judge: Gensyn’s verifiable AI evaluation system. Traditional evaluators rely on closed APIs - opaque, silently updated, and impossible to reproduce. Judge executes a pre-agreed, deterministic AI model against real-world inputs & commits to be challenged in

It was great talking with brilliant researchers on decentralised AI among many other things. Thanks for coming!

What is Gensyn?

Working on hard verification of ML jobs with @oguzer90 at the @gensynai offsite!

Gensyn community - we are hosting two separate events during ICML in Vancouver! Join us on July 15th & July 17th. Details 👇

1/ Introducing RL Swarm’s new backend: GenRL. A modular reinforcement learning library built for distributed, fault-tolerant training - now powering RL Swarm from the ground up. 🧵

Introducing CheckFree A fault tolerant method for decentralised training, with no checkpoints or redundant compute. Up to 1.6x faster than existing methods, with no convergence loss. We’re open sourcing it today.

Introducing NoLoCo NoLoCo trains large models over heterogeneous gossip networks, rather than high-bandwidth datacentres. It reduces synchronisation latency by 10x vs state of the art methods while converging 4% faster to the same validation loss. We're open sourcing it today.

Great to see the first Gensyn Community Meetup take place in Istanbul today, with our own @oguzer90 joining the panel. Thanks to @loop__creative for hosting!

Thank you for organizing the event! I also enjoyed participating and answering questions about Testnet and our research in general.

the MCDC workshop at ICLR from @Ar_Douillard et al. is a schelling point for decentralised AI last time I felt this was the NeurIPS 2022 Decentralization and Trustworthy ML in Web3 workshop from @quentinlou1, @dawnsongtweets, et al. vibe is different this time - we're growing

Come and find the @gensynai team at @encodeclub's AI London Hackathon this week -- hear more about our Testnet and take on our RL Swarm bounties!

Introducing the Gensyn Testnet The Gensyn testnet is live. Run a node, train your personal model, and track your participation in the swarm.

Introducing Diverse Expert Ensembles Diverse Expert Ensembles are a model architecture, optimised for decentralised networks. They are trained in an embarrassingly parallel fashion, and use expert diversity to enhance model performance.

Introducing SkipPipe SkipPipe is a new communication-efficient, pipeline-parallel training method. It reduces distributed training time by up to 55% and is scalable to theoretically infinite model size. Today, we're open sourcing it to push the frontier of decentralised ML.

Introducing Verde Verde is a high-scale verification system for machine learning over untrusted nodes. It ensures correct execution of ML operations across any device in the world, from data centres to the edge. It operates at runtime and doesn't require any onboarding.

The network for machine intelligence Two years ago, we laid out our vision for a machine learning compute protocol. One that connects every device in the world into an open network for machine intelligence, with no gatekeepers or artificial boundaries. This week, we’ll be