def setup_method(self, _):
from mlpy.mdp.stateaction import State, Action
State.nfeatures = None
Action.description = None
Action.nfeatures = None
case_template = {
"state": {
"type": "float",
"value": "data.state",
"is_index": True,
"retrieval_method": "knn",
"retrieval_method_params": 5
},
"act": {
"type": "float",
"value": "data.action",
"is_index": False,
"retrieval_method": "cosine",
},
"delta_state": {
"type": "float",
"value": "data.next_state - data.state",
"is_index": False,
}
}
from mlpy.knowledgerep.cbr.engine import CaseBase
self.cb = CaseBase(case_template, retention_method_params={'max_error': 1e-5})
from mlpy.auxiliary.io import load_from_file
self.data = load_from_file(os.path.join(os.getcwd(), 'tests', 'data/jointsAndActionsData.pkl'))
def main(args):
try:
data = load_from_file(args.infile)
obs = data["state"]
except IOError:
sys.exit(sys.exc_info()[1])
except KeyError, e:
sys.exit("Key not found: {0}".format(e))
def main(args):
try:
filename = convert_to_policy(args.policies)
# noinspection PyUnusedLocal
data = load_from_file(filename)
if ":" not in args.policy_num:
args.policy_num = args.policy_num + ":" + str(int(args.policy_num) + 1)
policies = eval("data['act'][" + str(args.policy_num) + "]")
except IOError:
sys.exit(sys.exc_info()[1])
except KeyError, e:
sys.exit("Key not found: {0}".format(e))
def main(args):
try:
data = load_from_file(args.infile)
train = data["train"]
test = data["test"]
obs_avg = calc_stats(test)
nobs = 20 # train[0].shape[0]
d, n = train[0][0].shape
except IOError:
sys.exit(sys.exc_info()[1])
except KeyError, e:
sys.exit("Key not found: {0}".format(e))
def setup_method(self, _):
from mlpy.mdp.stateaction import State, Action
State.nfeatures = None
Action.description = None
Action.nfeatures = None
case_template = {
"state": {
"type": "float",
"value": "data.state",
"is_index": True,
"retrieval_method": "knn",
"retrieval_method_params": 5
},
"act": {
"type": "float",
"value": "data.action",
"is_index": False,
"retrieval_method": "cosine",
},
"delta_state": {
"type": "float",
"value": "data.next_state - data.state",
"is_index": False,
}
}
from mlpy.mdp.continuous.casml import CASML
self.model = CASML(case_template, tau=1e-5, sigma=1e-5, ncomponents=2)
from mlpy.auxiliary.io import load_from_file
data = load_from_file(os.path.join(os.getcwd(), 'tests', 'data/jointsAndActionsData.pkl'))
# Extract 10th experience for testing
self.unseen_state = data["states"][0][:, 10]
self.unseen_action = data["actions"][0][:, 10]
self.model.fit(np.delete(data["states"][0], 10, 1), np.delete(data["actions"][0], 10, 1))
try:
data = load_from_file(args.infile)
train = data["train"]
test = data["test"]
obs_avg = calc_stats(test)
nobs = 20 # train[0].shape[0]
d, n = train[0][0].shape
except IOError:
sys.exit(sys.exc_info()[1])
except KeyError, e:
sys.exit("Key not found: {0}".format(e))
try:
data = load_from_file(args.policy)
actions = data["act"][args.policy_num]
except IOError:
sys.exit(sys.exc_info()[1])
except KeyError, e:
sys.exit("Key not found: {0}".format(e))
ntrials = 20
fig1 = None
fig2 = None
ax1 = None
ax2 = None
ax3 = None
ax4 = None
def test_agentmodule_creation(self):
from mlpy.agents.modules import AgentModuleFactory
# create follow policy module
with pytest.raises(TypeError):
AgentModuleFactory().create('followpolicymodule')
from mlpy.auxiliary.io import load_from_file
data = load_from_file(
os.path.join(os.getcwd(), 'tests', 'data/policies.pkl'))
with pytest.raises(AttributeError):
AgentModuleFactory().create('followpolicymodule', data)
AgentModuleFactory().create('followpolicymodule', data['act'][0:2])
# create learning module
with pytest.raises(TypeError):
AgentModuleFactory().create('learningmodule')
from mlpy.mdp.stateaction import Action
Action.set_description({
'out': {
'value': [-0.004]
},
'in': {
'value': [0.004]
},
'kick': {
'value': [-1.0]
}
})
# create `qlearner` learning module
AgentModuleFactory().create('learningmodule', 'qlearner', max_steps=10)
AgentModuleFactory().create('learningmodule',
'qlearner',
lambda s, a: 1.0,
max_steps=10)
with pytest.raises(ValueError):
AgentModuleFactory().create('learningmodule',
'qlearner',
1.0,
max_steps=10)
# create `rldtlearner` learner module
from mlpy.mdp.discrete import DiscreteModel
from mlpy.planners.discrete import ValueIteration
planner = ValueIteration(DiscreteModel(['out', 'in', 'kick']))
AgentModuleFactory().create('learningmodule',
'rldtlearner',
None,
planner,
max_steps=10)
AgentModuleFactory().create('learningmodule',
'rldtlearner',
planner=planner,
max_steps=10)
with pytest.raises(TypeError):
AgentModuleFactory().create('learningmodule',
'rldtlearner',
max_step=10)
explorer = None
if args.explorer_type in ["egreedyexplorer", "softmaxexplorer"]:
explorer = ExplorerFactory.create(args.explorer_type, args.explorer_params, args.decay)
if args.learner == "apprenticeshiplearner":
learner = None
if args.progress:
try:
learner = ApprenticeshipLearner.load(args.savefile)
except IOError:
pass
if not learner:
try:
data = load_from_file(args.infile)
obs = data["state"]
actions = data["act"]
labels = data["label"]
except IOError:
sys.exit(sys.exc_info()[1])
except KeyError, e:
sys.exit("Key not found: {0}".format(e))
# Train the model with empirical data
for i, (s, a, l) in enumerate(zip(obs, actions, labels)):
model.fit(s, a, l[0])
# model.print_transitions()
learner = ApprenticeshipLearner(
np.asarray(demo),
