7-1. Overview

Summary

EMETH is characterized by an architecture that not only generalizes and distributes various processes but also executes them at high speed by adopting a two-layer structure of Layer 1 and Layer 2. The EMETH L1/L2 servers, which form the core of this architecture, are composed of the following main components:

1. Splitter: Responsiblefor encrypting Jobs with confidentiality and dividing them into a format optimal for distributed processing. This allows the processing to be distributed across multiple nodes, reducing processing time.

2. Aggregator: Gathers and integrates learning-completed models in a distributed confidential encryption state. This enables the integration of results obtained through distributed processing and provides them as final output.

3. Verifier: A third party verifies whether the results of Jobs obtained through distributed processing are valid.

4. Signer: Signs tickets for Jobs (or SubJobs) processed in L2 at regular intervals. Each node can submit the ticket to EMETH Core on the L1 side Blockchain as soon as they receive it.

This architecture is not limited to language model training but has the flexibility to distribute various processes, enabling effective large-scale processing in a wide range of fields such as AI training and data analysis.

In the following sections, we will explain in detail the specific functions and roles of each component, as well as the overall processes of both L1 and L2.