Affine

About Affine

Affine is a project focused on synchronous inference-verification and training loops for Bittensor subnet120, enabling a process where model checkpoints are published, miner-contributed batches are privately submitted, and results are publicly audited and verified before advancing to the next training epoch. The system supports both mock and live chain adapters, emphasizing independent, auditable evidence checks for research epochs and operational states. Affine provides tools for miners, protocol checks, and GPU role verification, aiming to ensure transparent, reproducible, and secure model training and evaluation within the subnet.