mmlf¶
Module: missions.operations.mmlf
¶
Execute MMLF experiments
This module contains an operation and a process for MMLF. One MMLF process consists of a executing one run of a predefined number of episodes of a MMLF world. An MMLF operation consists of one or several runs of one or several MMLF worlds.
Note
The results of the runs are not stored in the standard operation result directories but in the MMLF ReadWrite Area. Only the outputs of the MMLF to stdout are stored in log files stored in the operation result directories.
Author: | Jan-Hendrik Metzen |
---|
Inheritance diagram for pySPACE.missions.operations.mmlf
:
Class Summary¶
MmlfOperation (processes, operation_spec, ...) |
Operation class for executing MMLF experiments. |
MMLFProcess (mmlf_path, world_conf_path, ...) |
Process for executing an MMLF run |
Classes¶
MmlfOperation
¶
-
class
pySPACE.missions.operations.mmlf.
MmlfOperation
(processes, operation_spec, result_directory, world_name, number_processes, create_process=None)[source]¶ Bases:
pySPACE.missions.operations.base.Operation
Operation class for executing MMLF experiments.
An MMLF operation consists of one or several runs of one or several MMLF worlds.
Class Components Summary
_createProcesses
(processes, world_pathes, ...)Function that creates the mmlf process. _flatten_dict
(input_dict)Flatten the (potentially) nested dict input_dict consolidate
()Consolidates the results of the MMLF operation. create
(operation_spec, result_directory[, ...])A factory method that creates an MMLF operation based on the -
__init__
(processes, operation_spec, result_directory, world_name, number_processes, create_process=None)[source]¶
-
classmethod
create
(operation_spec, result_directory, debug=False, input_paths=[])[source]¶ A factory method that creates an MMLF operation based on the information given in the operation specification operation_spec
-
MMLFProcess
¶
-
class
pySPACE.missions.operations.mmlf.
MMLFProcess
(mmlf_path, world_conf_path, learning_episodes, test_episodes, result_directory)[source]¶ Bases:
pySPACE.missions.operations.base.Process
Process for executing an MMLF run
One MMLF process consists of a executing one run of a predefined number of episodes of a MMLF world.
The following parameters are mandatory:
mmlf_path: The path to the MMLF code world_conf_path: The relative path of the world configuration file within the specified MMLF rw area learning_episodes: The number of episodes the MMLF agent is allowed to train (i.e. optimize its policy) test_episodes: The number of episodes the learned policy is tested. This can be set to 1 for deterministic environments. result_directory: The directory where the stdout of the MMLF run is stored Class Components Summary
__call__
()Executes this process on the respective modality