Kup Compute Horde (SN12) w Canada

Compute Horde, identified as Subnet 12 within the Bittensor network, is a decentralized infrastructure project designed to provide scalable and trusted GPU computing power. Its primary objective is to serve as a high-performance compute layer for the entire Bittensor ecosystem, allowing validators of other subnets to offload resource-intensive tasks to a distributed pool of hardware. By doing so, the project aims to reduce the reliance of decentralized AI networks on centralized cloud service providers. The technical architecture of Compute Horde introduces several innovations to the Bittensor ecosystem. One of its main features is the use of executors, which allows individual miners to manage multiple hardware units simultaneously. This design enables the subnet to scale its total available computing power far beyond the standard limitations of the network. The project categorizes hardware into specific classes, ensuring that computational tasks are matched with appropriate resources, such as specific GPU models required for complex machine learning jobs. To maintain integrity and fairness, Compute Horde employs specialized validation mechanisms. It was one of the first subnets to implement a commit-reveal scheme, which prevents malicious actors from simply copying the work or scores of others. This system ensures that miners provide authentic proof of work. Additionally, the project uses a facilitator component that acts as an interface for organic job requests, allowing external users and other subnets to submit tasks directly to the pool. Developed by a team associated with Backend Developers Ltd, Compute Horde supports a wide range of use cases beyond AI model training, including video processing and hosting open-source models from platforms like Hugging Face. By providing a software development kit, the project enables other developers to integrate decentralized GPU resources into their own applications easily. Ultimately, Compute Horde seeks to establish a foundational infrastructure that supports the long-term scalability of decentralized intelligence and general-purpose computing. Dowiedz się więcej