Skip to main content

Hyper Parameters

Each model directory contains a hyper_parameters.json file: a JSON object of training and inference settings that are specific to that model (circuit, system, or task). The ADK loads this file into GenieModel.hyperparameters at run time.

On disk​

{
  "learning_rate": 0.001,
  "batch_size": 64
}

There are no enforced keys — structure and interpretation are entirely up to your agent implementation. Read them from agent_data.optimization_data.genie_model.hyperparameters (RL path) or ctx.optimization_data.genie_model.hyperparameters (custom executor path).

Python type​

hyperparameters: dict[str, Any]

Part of GenieModel. Values are also included in platform transfer payloads when creating a child model; see Model handling.

  • Models — creating models with genie model add
  • GenieModel — full model structure loaded from disk