amaze.extensions.sb3

Contains wrapper code to make AMaze work smoothly with stable baselines 3

Functions

compatible_models()	Returns the list of SB3 models that can be used with this extension
load_sb3_controller(path)	Loads a wrapper stable baselines 3 model from an archive.
sb3_controller(robot_data, model_type, ...)	Creates a wrapper for a stable baselines 3 model.

amaze.extensions.sb3.compatible_models()

Returns the list of SB3 models that can be used with this extension

amaze.extensions.sb3.sb3_controller(

robot_data: BuildData,

model_type: Type[BaseAlgorithm],

*args,

**kwargs,

)

Creates a wrapper for a stable baselines 3 model.

Parameters:

• robot_data – Input and output spaces specifications

• model_type – the type of SB3 model to wrap (PPO, A2C, …)

• args – positional arguments to pass to the model

• kwargs – keyword arguments to pass to the model

amaze.extensions.sb3.load_sb3_controller(path: str | Path)

Loads a wrapper stable baselines 3 model from an archive.

Warning:

do not forget to specify use of this extension (sb3) when loading from an executable from the core library (e.g. main)

Modules

amaze.extensions.sb3.callbacks	Contains an out-of-the-box example of verbose callback relying on Tensorboard.
amaze.extensions.sb3.controller	Implements a wrapper around common models from stable baselines 3
amaze.extensions.sb3.graph	Collection of functions to generate a user-friendly representation of a neural network from stable baselines 3
amaze.extensions.sb3.guard	Implementation of a python guard to prevent negative interactions between opencv2 and PyQT5
amaze.extensions.sb3.maze_env	SB3 wrapper for the maze environment
amaze.extensions.sb3.networks	Half-hearted attempt a making a custom CNN.
amaze.extensions.sb3.utils	Various utility functions