Skip to content

AI_Core_Training_Module.md

Latest commit

 

History

History
69 lines (38 loc) · 3.8 KB

AI_Core_Training_Module.md

File metadata and controls

69 lines (38 loc) · 3.8 KB
Raw
graph LR
    AI_Training_Orchestrator["AI Training Orchestrator"]
    Core_AI_Algorithm_Engine["Core AI Algorithm Engine"]
    Distributed_Training_Coordinator["Distributed Training Coordinator"]
    Distributed_Training_Worker["Distributed Training Worker"]
    Single_Process_Training_Handler["Single-Process Training Handler"]
    AI_Training_Orchestrator -- "initiates" --> Single_Process_Training_Handler
    AI_Training_Orchestrator -- "initiates" --> Distributed_Training_Coordinator
    Distributed_Training_Coordinator -- "assigns tasks to" --> Distributed_Training_Worker
    Distributed_Training_Worker -- "reports to" --> Distributed_Training_Coordinator
    Distributed_Training_Worker -- "utilizes" --> Core_AI_Algorithm_Engine
    Single_Process_Training_Handler -- "calls" --> Core_AI_Algorithm_Engine
Loading

CodeBoardingDemoContact

Details

The AI Core & Training Module subsystem, encapsulated within the poker_ai.ai package, manages the AI training lifecycle, including core algorithms, distributed training, and single-process training.

AI Training Orchestrator

Serves as the primary interface for the AI training lifecycle, responsible for initiating and resuming training processes. It dispatches training requests to either single-process or distributed handlers based on configuration.

Related Classes/Methods:

Core AI Algorithm Engine

Implements the fundamental AI algorithms, specifically Counterfactual Regret Minimization (CFR) and CFR+. This component is responsible for the iterative calculation and updating of the AI's strategy, forming the computational core of the learning process.

Related Classes/Methods:

Distributed Training Coordinator

Orchestrates and manages distributed AI training. Its responsibilities include starting worker processes, distributing training jobs, and synchronizing their progress and results across multiple nodes or processes.

Related Classes/Methods:

Distributed Training Worker

Executes individual AI training tasks within a distributed setup. Each worker performs iterations of the core AI algorithms and updates its local strategy, communicating with the Distributed Training Coordinator for task assignment and result reporting.

Related Classes/Methods:

Single-Process Training Handler

Manages the AI training process when executed in a single-threaded, non-distributed environment. It directly invokes the Core AI Algorithm Engine to perform training iterations without the overhead of distributed coordination.

Related Classes/Methods:

FAQ