🍃 Lilypad Project Report: October 24, 2024
v2.8.0 release, Lilypad Dell Quantum Computing hackathon, and more!
🌎 Overview
The Lilypad IncentiveNet continues to grow, aggregating a diverse range of GPUs from GTX 3070s to A100s! This compute marketplace would not be possible without the global community of Lilypad Resource Providers (RPs). RPs run GPU nodes and process jobs from the network and running a Proof of Work (PoW) throughout the day to report uptime and GPU card type. RPs are able to monetize their idle GPU capacity with demand coming from Lilypad!
Thank you to the early Lilypad Resource Provider community! Check out the Lilypad metrics page to view a global Lilypad node map.
Lilypad's compute capacity is sourced from around the globe, providing a platform to run AI models and other arbitrary programs with low-cost, high-performance compute.
Read more in the docs about creating a custom Lilypad module that can run as a job on the network! Developers and ML engineers leverage the Lilypad CLI or smart contracts to run jobs on the network. As the Lilypad economy grows, our team will provide a module marketplace SaaS platform to build and deploy a custom Lilypad module easily.
Run a Lilypad node , check out our developer docs, and join us on Discord in this exciting journey towards a more distributed and accessible AI ecosystem!
⚒️ Engineering Update
The Lilypad core engineering team has been heads down planning, building, and testing systems for Lilypad Mainnet. This includes working in collaboration with our Research team to review the Lilypad Whitepaper and Network Tokenomics. Learnings from the IncentiveNet have been critical to help determine the right path forward for Mainnet. The job economy, RP proofing mechanisms, multi-chain architecture, and more have all been inspired by the learnings from IncentiveNet. In addition, our team continues to build out relationships with key partners to provide added layers of value including large scale GPU clusters for large compute jobs (over 24GB of VRAM required), privacy systems, decentralized data storage, and more!
IncentiveNet and the Lilybit_ rewards program provide a production environment for running Lilypad jobs and optimizing Network orchestration.The PoW rewards mechanism, job matching engine (Solver), and more have all been tested under the pressure of over 2,000 Lilypad nodes! Our team continues to push updates that help improve job performance as well as RP interactions with the Network.
RPs update to the latest version! - v2.8.0 Dijkstra
"Simplicity is prerequisite for reliability." ~ Edsger Dijkstra
Lilypad v2.8.0 - Dijkstra was released, requiring Lilypad RPs to update! V2.8.0 focuses on improving the consistency of jobs running on the network.
RPs must have a minimum balance of .0006 Arbitrum ETH for resource offers to be accepted to run Lilypad jobs.
RPs now need to run a local ipfs node. Running a local ipfs node improves the reliability and speed of running/completing jobs on the network.
After a job run, RPs will now re-register resource offers with the Solver.
For more details, check out the release notes.
Local ipfs node on Lilypad RPs
A change implemented in v2.8.0 includes requiring RPs to set up a local ipfs node on their Lilypad RP. When a job runs successfully on Lilypad, the output is saved in a local folder on the client as well as on ipfs. This ipfs system was using an ipfs node embedded in Bacalhau (used by the Lilypad RP). The embedded ipfs node was causing performance problems when returning results to end users.
To address this problem, our team updated the Lilypad RP docs to include info regarding how to setup a local ipfs node. RPs will install Kubo (go-ipfs), configure their Bacalhau systemd file to use the local node instead of the embedded ipfs node, and then initialize the daemon. Once the ipfs node is running, RPs can then start Bacalhau and the Lilypad-Resource-Provider. Thank you to RPs supporting this update. Consistency of jobs running on the Lilypad Network continue to improve!
Follow the updates-rp channel in the Lilypad Discord to keep up to date with the action!
⚡ Developer Events & DX Updates
Community dashboards
The Lilypad ecosystem uses several public dashboards for network transparency and monitoring Lilybit_ rewards earned.
The Lilypad Leaderboard displays a count of total Lilybit_s earned by each RP. Nodes can also view the status of their RP in the “Node Status” tab. The “Metrics” tab shows a global node map of the Lilypad Network.
The RP-points dashboard is a tool for RPs to view Lilybit_ reward earnings by viewing daily earnings data. This data gives RPs a more detailed view into their rewards and is helpful for RP troubleshooting.
The Open Source + Ambassadors dashboard provides a public list of all community rewards earned by Lilypad contributors. From open source contributions to serving as a Lilypad Ambassador, the Lilybit_s earned for each group are updated at the end of each month!
Hacktoberfest
Contribute to the Open Source fun by getting involved in Hactoberfest! Running until the end of October, jump into the Lilypad main repo and look for issues marked at “Community” and “Hacktoberfest”. Completed issues are counted in your Hactoberfest monthly contribution tally!
Weekly AMA Sessions!
To help answer questions and gain insights from our team, Lilypad is hosting AMA sessions every Wednesday! We will alternate the timing of the AMA every week to ensure we can cater to RPs and devs across the globe! Find an AMA time that works for you on the Lilypad Calendar.
Catch up with the AMA from last week with DeveloperSteve and Taosheng!
🌟Community
Lilypad <> Dell Quantum Computing Hackathon
Our team was thrilled to co-host a Quantum Computing hackathon in collaboration with Dell Technologies at LA Tech Week! This event brought together researchers focused on Quantum algorithm research using systems like cuQuantum! Lilypad is super proud to provide compute for these types of use cases and will continue to support researchers looking for compute to run simulation circuits like cuQuantum.
A new era of Enzyme engineering - Amelie Schreiber
The BioML community continues to make great strides in rare disease research, enzyme/protein creation, and, ultimately, drug discovery.
Amelie Schreiber, a researcher pushing the boundaries of what’s possible with BioML, recently published a blog on Huggingface discussing “A new era of enzyme engineering”. In this piece, Amelie discusses the workflow of using models like RFDiffusion with other leading BioML models to generate new enzymes for medical and commercial applications. The application of Large Language Models(LLMs) and other AI techniques to medical research is a significant use case supported by the Lilypad Network and Module Marketplace. As the Lilypad Network and tools set grows, we are proud to provide the system needed to help accelerate this field or research!
Lilypad at the Chan Zuckerberg Initiative
Lilypad Research Lead ,ScienceStanley presented at the Chan Zuckerberg Initiative about the crucial work of building BioML model pipelines to solve complex problems like rare disease research. As mentioned above, the Lilypad Network team is excited to continue providing low-cost, high-performance computing for both researchers and engineers in the field of BioML.
In addition to supporting this use case today, the Module Marketplace platform launching in early 2025 will include tooling geared towards simplifying the process of using Lilypad compute in these workflows. The platform will offer a simple interface to run a model on Lilypad and then interact with the model with our SDK.
To learn more about the latest R&D developments happening on the Lilypad Network, tune into our weekly X Spaces! Last week, the Lilypad team was joined by Conduit to discuss Trusted Execution Environments and running private jobs on Lilypad. Follow Lilypad on X to catch upcoming spaces!
Proof of Talk Podcast
Lilypad CEO Ally Haire and Head of Research Stanley Bishop recently made waves on the Proof of Talk podcast, discussing their unconventional journeys into decentralized computing. Ally, a former cafe owner turned tech leader, and Stanley, a bioinformatician dedicated to open science, shared the driving force behind Lilypad: empowering researchers with affordable, accessible compute power. They hopped into the platform's unique architecture, its potential to revolutionize fields like genomics research, and the exciting progress of the IncentiveNet leading to Mainnet. Tune in to discover how Lilypad is shaping the future of AI and decentralized technology!
🔮 What's Next?
Our team continues to push improvements to the Lilypad IncentiveNet focused on job consistency and smooth scaling of the job matching system. The Lilypad Network intends to support many different GPU architectures in addition to Nvidia. Our team is currently working on adding support for AMD GPU architecture with ROCm drivers. There’s a major opportunity to support running AI models on AMD cards as many individuals and businesses are struggling with the complexities and issues around running parallel processes using ROCm. The Lilypad team is super excited to support multi-GPU architectures providing a software layer to abstract this complexity away.
☎️ Contact Us
💬 Have more questions or looking for guidance? Join our community AMA hosted every Wednesday!
Chat to us on Discord and follow along on X for the latest news and updates!