Current Competitions

Subnet 1 Apex competitions registry

Matrix Compression

The first competition - Matrix Compression - explores how small neural activations - both forward and backward - can be compressed while still retaining all their original information. Reducing activation size enables faster data transfer across the internet, a crucial step toward making distributed training more efficient, as it’s often constrained by network bandwidth. The top-performing algorithms from this competition will be integrated to enhance training on subnet 9 IOTA.

Miners aim to optimize the following:

  • Compression Ratio - How small the compressed solution is on disk versus the starting matrix.

  • Time - How fast the compression/decompression algorithm runs.

To surpass the current winner of the competition, a miner must earn a raw score of at least 1% higher than the current top score. If there is no current winner, then a miner must earn a raw score of at least 1% higher than the baseline score.

  • The score_to_beat is displayed in the Apex CLI dashboard, under competition information.

For Miners

View the matrix compression baseline miner solution provided as an example.

View the matrix compression general miner solution template.

Then, continue to the Apex CLI guide to submit a solution.

Overview

Why Matrix Compression matters so much in decentralized LLM training:

  1. Activations must be communicated or stored between layers/devices. Without compression, this inter-layer communication becomes the major bottleneck in throughput and scalability.

  2. Matrix Compression enables effective scalability across many nodes by mitigating network communication latencies. 

  3. Reducing the memory footprint of activations frees up resources for other parts of the pipeline, improving the overall efficiency and cost-effectiveness of decentralized training.

Innovative matrix compression solutions unlock several high-impact real-world applications:

  • Distributed AI and Federated Learning Efficiency

  • Data Center and Cloud Infrastructure Optimisation

  • Edge and Networked Systems Performance

  • Large-Scale Simulation and Optimisation Systems.

Optimisation advancements translate into lower network strain in AI-driven systems, more sustainable infrastructure, faster, scalable collaboration between machines and data centres, real-time enhancement for logistics, transportation, and communication networks.

Last updated