bam.logs

Classes

Logs	Collection of processed trajectory logs used for identification.

Module Contents

class bam.logs.Logs(directory: str)

Collection of processed trajectory logs used for identification.

Loads all JSON files found in a directory (produced by python -m bam.process) and exposes them as a list of log dicts. Each log dict contains the pendulum configuration (mass, arm_mass, length, kp, vin) and a list of timestep entries with position, velocity, and control values.

Parameters:

directory – Path to a directory of processed JSON log files.

directory: str

json_files

logs = []

split(selector_kp: int) → Logs

Split logs by P-gain value to create a validation set.

Removes all logs recorded with kp == selector_kp from this object and returns them as a new Logs instance. Modifies self in place.

Parameters:

selector_kp – P-gain value used to select the held-out logs.

Returns:

A new Logs object containing only the held-out logs.

make_batch() → dict

Make a batch log from all the logs. In a batch log, all entries are vectorized. For example, batch[“mass”] is a vector of all masses batch[“entries”][0][“position”] will be a vector of all positions