Python API for Brain Simulator

Python module for using GoodAI's Brain Simulator

The module itself serves as an interface to Brain Simulator's MyProjectRunner. You can find a documentation to MyProjectRunner on GoodAI's docs website.

Install from source

• Go to folder with setup.py

• Run pip install . or pip install . --upgrade if you are upgrading

Note: GoodAI-BrainSim module depends on Python for .NET which cannot be installed using easy_install, therefore you cannot use the usual python setup.py install

How to use the module

• First thing you need to do is import a load_brainsim method from the module
• Then you have to call this method. It has one optional argument in which you can specify a path to the folder where you have Brain Simulator installed. If you do not provide the path, it will try to find a path itself in the Windows registry. If the method is unable to find a path in registry, it will use the path C:\Program Files\GoodAI\Brain Simulator.
• After that, you can import other types from the loaded Brain Simulator modules. All classes except MyProjectRunner are imported from the corresponding Brain Simulator namespaces
  • You will always want to import MyProjectRunner. It is used to control the Brain Simulator. It is imported from the goodai.brainsim package
  • Other classes are imported from GoodAI namespaces. E.g. from GoodAI.Core.Utils import MyLog
  • If you want to use Brain Simulator's School for AI, you want to import SchoolCurriculum, CurriculumManager and LearningTaskFactory as well
  • You can also import some other types useful e.g. for curriculum creation or logging the events. Those are ToyWorldAdapterWorld and MyLog

Together, beginning of your script may look like

from goodai.brainsim import load_brainsim
load_brainsim("C:\Users\john.doe\my\brainsimulator\path")

from goodai.brainsim import MyProjectRunner
from GoodAI.Modules.School.Common import SchoolCurriculum, LearningTaskFactory
from GoodAI.School.Common import CurriculumManager
from GoodAI.School.Worlds import ToyWorldAdapterWorld

# your code here

If you want to know how to use MyProjectRunner please refer to its documentation.

Example usage

You can find a simple example of using the API in examples/basic.py or below. It creates simple curriculum from tasks for ToyWorldAdapterWorld and then runs the curriculum for 10 steps sending random actions to Brain Simulator.

import random
from goodai.brainsim import load_brainsim
load_brainsim()

from goodai.brainsim import MyProjectRunner

from GoodAI.Modules.School.Common import SchoolCurriculum, LearningTaskFactory
from GoodAI.School.Common import CurriculumManager
from GoodAI.School.Worlds import ToyWorldAdapterWorld

# Get Node
# node = runner.GetNode(22)
# or
# node = runner[22]

# Get Memory Block
# memblock = runner[22].Visual

# Get Values
# floatField = runner.GetValues(22, "Visual")
# or
# memblock = runner[22].Visual
# memblock.SafeCopyToHost()
# floatField = memblock.Host


runner = MyProjectRunner()
runner.OpenProject('C:/Users/john.doe/Desktop/test.brain')
runner.DumpNodes()

school = runner[22]

curr = SchoolCurriculum()

for w in CurriculumManager.GetAvailableWorlds():
    if w.Name == ToyWorldAdapterWorld.__name__:
        for t in CurriculumManager.GetTasksForWorld(w):
            it = LearningTaskFactory.CreateLearningTask(t, school)
            it.RequiredWorldType = w
            curr.Add(it)

school.Curriculum = curr

actions = 13 * [0]

runner.RunAndPause(1)


for _ in xrange(10):
    actions[0] = random.random()
    print(actions)
    runner.SetValues(24, actions, "Output")
    runner.RunAndPause(1)
    tWorld = school.CurrentWorld
    chosenActions = tWorld.ChosenActions
    tWorld.ChosenActions.SafeCopyToHost()
    print(chosenActions.Host[0])

runner.Shutdown()

Other examples can be found in test folder.

In test/test_school.py you can see simple usage of custom MyTextLogWriter, TrainingResult and loading saved curricula.