Configuring the Inference Pipeline

Pipeline order: observation → preprocessors → runner → postprocessors → action output. See the pre/post-processing guide and manifest schema.

Workflow

1. Read the manifest slice for preprocessors, postprocessors, and model.runner.

   • Done when: you know whether specs use type (registry short name) or class_path.

2. Prefer type for built-ins registered by your installed Runtime package (e.g. normalize/denormalize patterns in the docs).

3. Use class_path + init_args for explicit classes:

   preprocessors:
     - class_path: physicalai.inference.preprocessors.StatsNormalizer
       init_args:
         artifact: stats.safetensors
   

   • Done when: init_args paths resolve relative to the export directory via resolve_artifact.