Common Configuration

Similar to the original GS, Splatwizard’s configuration architecture is broadly divided into three modules: pipeline configuration PipelineParams, model configuration ModelParams, and optimization parameter configuration OptimizationParams. The pipeline configuration includes a series of common settings, such as dataset paths, output paths, runtime modes, and so on. The model configuration and optimization parameter configuration can be customized according to the characteristics of each model. Accordingly, it is also necessary to register the corresponding model parameter class and optimization parameter class for each model. Splatwizard uses simple-parsing as the foundational configuration tool, so all configuration items are represented in the code as dataclasses.

Note if you want to define your model configuration and optimization parameter configuration, you should inherit ModelParams and OptimizationParams

from dataclasses import dataclass
from splatwizard.config import ModelParams, OptimizationParams


@dataclass
class GSModelParams(ModelParams):
    pass


@dataclass
class GSOptimizationParams(OptimizationParams):
    pass