Lumino is a decentralized machine learning training protocol that connects compute providers with model training demanders via a blockchain network. It allows compute providers (such as data centers or individual GPUs) to connect idle resources, complete training tasks, and earn token rewards; model developers can submit training jobs, configure parameters (e.g., batch size, learning rate, epochs, LoRA), and execute them in distributed or multi-GPU environments. Users manage accounts, stake tokens, create and run jobs, and monitor network status through a command-line client (go-client). Its mechanism uses smart contracts to coordinate supply and demand, aiming to reduce training costs and improve resource utilization.