CLI: Infer

Preview: The CLI is a planned API. The examples below document the target design.

Run inference from a config or exported policy package.

physicalai infer --config inference.yaml

Example config:

model:
  class_path: physicalai.inference.InferenceModel
  init_args:
    export_dir: ./exports/act_policy
    backend: openvino
    device: CPU

The Python equivalent:

model = InferenceModel.load("./exports/act_policy")
action = model.select_action(observation)

Use physicalai run for robot control loops. Use physicalai infer for offline inference or testing outside a runtime loop.