Join Our Collaborative Experiment to Distribute Large Language Models Across Multiple Machines
We are seeking enthusiastic participants to assist with an innovative experiment in distributed machine learning. Our goal is to explore methods for efficiently splitting and running the GLM 5.2 language model across multiple hardware nodes.
Project Overview
Our team has developed a custom network protocol designed to distribute neural network weights and coordinate activation processes across different machines. This approach aims to facilitate usable inference by leveraging multiple nodes working in concert. Currently, we have a working prototype with two nodes, and we are eager to expand this setup to larger configurations.
What We’re Looking For
We are particularly interested in assembling a small team of eight individuals, each equipped with a machine that has at least 64GB of RAM—preferably MacBook M5 Pros for hardware homogeneity, as we already have two in this configuration. However, we are open to participating hardware and are happy to include a variety of systems willing to collaborate.
If you are interested in contributing your resources to this experiment, please reach out by commenting with your specifications. We will coordinate further through our dedicated Discord server to facilitate smooth collaboration.
Why Participate?
This project is driven by academic curiosity and aims to refine hardware utilization techniques to improve efficiency in deploying large language models. Our current setup involves two nodes, but scalability to ten or more nodes presents significant challenges and opportunities, particularly around network bandwidth constraints.
Current Progress and Future Goals
While previous efforts in this domain have faced limitations, we are optimistic about the potential improvements and novel approaches we are exploring. Your participation could contribute valuable insights into addressing these challenges.
Next Steps
We appreciate the enthusiastic community engagement and look forward to collaborating with participants eager to push the boundaries of distributed machine learning. Stay tuned for updates as we gather data and analyze results.
Contact and Collaboration
If you’re interested in helping us validate these techniques, please comment below with your system details. Join our Discord community for ongoing discussions and coordination.
Thank you for your interest and support in advancing this exciting area of AI research.
Note: All experiments adhere to ethical guidelines and aim to contribute to the open scientific community.
