def example():
name = "Hello"
ctor = Method("Hello", [
Argument("double", "dummy"),
Argument("double", "x", 1.0),
Argument("int", "j", 1),
Argument("char *", "name", "hee"),
Argument("char *", "password", 0),
], ["std::cout << x << std::endl;", "m_x = x;", "m_arr = new double[10];"])
dtor = Method("~Hello", [], ["delete [] m_arr;"])
call = Method("operator()", [Argument("char *", "someone")],
['''std::cout << "hello " << someone << std::endl;'''],
type="void")
m_x = Member("double", "m_x")
m_arr = Member("double *", "m_arr")
parentClass = Class('Parent')
klass = Class(name,
public_methods=[ctor, dtor, call],
private_members=[m_x, m_arr],
namespace="Greeting::Goodwill",
headers_dependent_on=["iostream"],
parents=[parentClass])
assert klass.constructors() == [ctor]
return klass
def __init__(self, spec):
Method.__init__(self)
plan = ExperimentPlan(Context.SimulationChain)
print "MethodDOE: ", spec
plan.Import(open(spec["Plan"]))
self.Plan = plan
#self._Event = threading.Event()
#self._Event.set()
self._Sem = threading.Semaphore(0)
def test(self):
method = Method()
method.id = 'a'
method.startTime = -1.
method.equilibrationTime = 1.
method.recordingTime = 2.
method.maximumSteps = 1e9
method.numberOfFrames = 100
method.solverParameter = 0.01
stream = StringIO()
writeMethodXml(method, stream)
def __init__(self, spec):
Method.__init__(self)
plan = ExperimentPlan(Context.SimulationChain)
print "MethodDOE: ", spec
plan.Import(open(spec["Plan"]))
self.Plan = plan
if spec.has_key("InfillCriteriaFunction"):
self.InfillCriteriaFunction = spec["InfillCriteriaFunction"]
else:
self.InfillCriteriaFunction = AdaptiveDOECriteria.SimpleLOO
self.MaxIterations = spec["Iterations"]
self.NewSitesPerIteration = spec["NewSitesPerIteration"]
self._Sem = threading.Semaphore(0)
def test_execute_void_no_parameters(self):
from VM import VM
vm = VM()
m = Method()
m.name = 'TestMethod'
m.returnType = Types.Void
m.parameters = []
vm.methods.append(m)
self.assertEqual(vm.current_method(), None)
vm.execute_method(m)
self.assertEqual(vm.current_method(), m)
r = Ret()
r.execute(vm)
self.assertEqual(vm.current_method(), None)
def test_execute_void_no_parameters(self):
from VM import VM
vm = VM()
m = Method()
m.name = 'TestMethod'
m.returnType = Types.Void
m.parameters = []
vm.methods.append(m)
self.assertEqual(vm.current_method(), None)
vm.execute_method(m)
self.assertEqual(vm.current_method(), m)
r = Ret()
r.execute(vm)
self.assertEqual(vm.current_method(), None)
def get_method(self):
m = Method()
m.methodDefinition = self;
m.locals = copy.deepcopy(self.locals)
#m.parameters= copy.deepcopy(self.parameters)
m.instructions = self.instructions
m.maxStack = self.maxStack
m.returnType = self.returnType
m.attributes = self.attributes
return m
def __init__(self, spec):
Method.__init__(self)
self.Chromosome = Factory.Chromosome(Context.SimulationChain)
self.ResEval = ChainResponseEvaluator()
self.ResEval.Start()
self.PenEval = Factory.CreatePenaltyEvaluator(spec)
self.FitnessEval = EA.FitnessEvaluator()
optiConfig = spec["OptimizerConfig"]
popSize = optiConfig["PopulationSize"]
self.OptimizerConfig = optiConfig
postIterCallback = lambda gen, pop: self.PostIteration(gen, pop)
self.BestExperiment = None
if spec["Optimizer"] == "GA":
self.Optimizer = EA.MonoObjectiveGA(self.Chromosome, self.ResEval,
self.PenEval, self.FitnessEval,
popSize, 2)
self.Optimizer.PostIterationCallback = postIterCallback
self.HistoryFile = open("%s.history.dat" % Context.Name, "w")
elif spec["Optimizer"] == "SurrogateGA":
Context.Database.Import(open(optiConfig["DOE"]))
pop = Factory.PopulationFromPlan(self.Chromosome, Context.Database)
self.Optimizer = EA.MonoObjectiveSurrogateGA(
self.Chromosome, self.ResEval, self.PenEval, self.FitnessEval,
pop)
elif spec["Optimizer"] == "ParetoGA":
self.Optimizer = EA.SPEA2(self.Chromosome, self.ResEval,
self.PenEval,
optiConfig["PopulationSize"],
optiConfig["ArchiveSize"])
elif spec["Optimizer"] == "ParetoSurrogateGA":
Context.Database.Import(open(optiConfig["DOE"]))
pop = Factory.PopulationFromPlan(self.Chromosome, Context.Database)
self.Optimizer = EA.ParetoSurrogateGA(
self.Chromosome, self.ResEval, self.PenEval, pop, popSize,
optiConfig["ArchiveSize"], optiConfig["Update"],
optiConfig["UpdateTruncation"])
def test_execute_instance_clears_stack_with_void_return_type(self):
from VM import VM
vm = VM()
m = Method()
m.name = 'TestMethod'
m.attributes.append(MethodDefinition.AttributeTypes['instance'])
m.returnType = Types.Void
m.maxStack = 99
m.parameters = []
vm.methods.append(m)
vm.stack.push(111)
vm.execute_method(m)
vm.stack.push(124)
vm.stack.push(987)
r = Ret()
r.execute(vm)
self.assertEqual(vm.current_method(), None)
self.assertEqual(vm.stack.count(), 1)
self.assertEqual(vm.stack.pop(), 111)
def test_execute_int_no_parameters_returns_value_on_stack(self):
from VM import VM
vm = VM()
m = Method()
m.name = 'TestMethod'
m.returnType = Types.Int32
m.parameters = []
vm.methods.append(m)
v = Variable(888)
self.assertEqual(vm.current_method(), None)
self.assertEqual(vm.stack.get_frame_count(), 0)
vm.execute_method(m)
vm.stack.push(v)
self.assertEqual(vm.current_method(), m)
# fixme - test return value too
r = Ret()
r.execute(vm)
self.assertEqual(vm.current_method(), None)
self.assertEqual(vm.stack.get_frame_count(), 1)
self.assertEqual(vm.stack.pop(), v)
def test_execute_int_no_parameters_returns_value_on_stack(self):
from VM import VM
vm = VM()
m = Method()
m.name = 'TestMethod'
m.returnType = Types.Int32
m.parameters = []
vm.methods.append(m)
v = Variable(888)
self.assertEqual(vm.current_method(), None)
self.assertEqual(vm.stack.get_frame_count(), 0)
vm.execute_method(m)
vm.stack.push(v)
self.assertEqual(vm.current_method(), m)
# fixme - test return value too
r = Ret()
r.execute(vm)
self.assertEqual(vm.current_method(), None)
self.assertEqual(vm.stack.get_frame_count(), 1)
self.assertEqual(vm.stack.pop(), v)
def testTimeInhomogeneous(self):
model = StateModel()
model.id = 'model'
model.speciesIdentifiers.append('s1')
model.species['s1'] = StateSpecies('', 'species 1', '13')
model.speciesIdentifiers.append('s2')
model.species['s2'] = StateSpecies('', 'species 2', '17')
model.reactions.append(
StateReaction('r1', 'reaction 1', [StateSpeciesReference('s1')],
[StateSpeciesReference('s2')], True, '2+sin(t)'))
model.reactions.append(
StateReaction(
'r2', 'reaction 2',
[StateSpeciesReference('s1'),
StateSpeciesReference('s2')],
[StateSpeciesReference('s1', 2)], False, '1+exp(-t)'))
error = model.evaluateInhomogeneous()
assert not error
method = Method()
output = TimeSeriesFrames()
solver = FirstReaction(makeModel(model, method), 2**64)
simulator = makeTimeSeriesUniform(solver, model, output)
def test_execute_instance_clears_stack_with_void_return_type(self):
from VM import VM
vm = VM()
m = Method()
m.name = 'TestMethod'
m.attributes.append(MethodDefinition.AttributeTypes['instance'])
m.returnType = Types.Void
m.maxStack = 99
m.parameters = []
vm.methods.append(m)
vm.stack.push(111)
vm.execute_method(m)
vm.stack.push(124)
vm.stack.push(987)
r = Ret()
r.execute(vm)
self.assertEqual(vm.current_method(), None)
self.assertEqual(vm.stack.count(), 1)
self.assertEqual(vm.stack.pop(), 111)
from Dataset import Dataset
from Method import Method
from Evaluation import Evaluation
from Result import Result
from Setting import Setting
import numpy as np
#--------------------------- run the exp ----------------------------
if 1:
k = 5
fold = 10
dataset = Dataset('', '')
dataset.file_folder_path = '../data/input/'
method = Method('', '')
method.k = k
evaluation = Evaluation('')
result = Result('', '')
result.k = k
setting = Setting('', '', dataset, method, result, evaluation)
setting.fold = fold
setting.load_classify_save()
if 1:
fold = 10
k = 5
result = Result('', '')
def main():
from Species import Species
from Method import Method
from SpeciesReference import SpeciesReference
print('-' * 79)
print('Time Homogeneous')
model = Model()
model.id = 'model'
model.compartments['C1'] = Value('Cell', '1')
model.speciesIdentifiers.append('s1')
model.species['s1'] = Species('C1', 'species 1', '13')
model.speciesIdentifiers.append('s2')
model.species['s2'] = Species('C1', 'species 2', '17')
model.reactions.append(
Reaction('r1', 'reaction 1', [SpeciesReference('s1')],
[SpeciesReference('s2')], True, '1.5'))
model.reactions.append(
Reaction('r2', 'reaction 2',
[SpeciesReference('s1'),
SpeciesReference('s2')], [SpeciesReference('s1', 2)], True,
'2.5'))
writeModelXml(model)
print('')
print(model.writeSpeciesTable())
print('')
print(model.writeReactionsTable())
# The model must be evaluated for writing SBML or the C++ code.
model.evaluate()
writeModelSbml(model, open('model.xml', 'w'))
method = Method()
writeModelMathematica(model, method, open('model.nb', 'w'))
print('')
print(model.makePropensitiesNumberOfReactions())
print(model.makePropensitiesConstructor())
print(model.makePropensitiesMemberFunctions(True))
print('')
model.writeCmdl(sys.stdout)
print('-' * 79)
print('Time Inhomogeneous')
model = Model()
model.id = 'model'
model.speciesIdentifiers.append('s1')
model.species['s1'] = Species('', 'species 1', '13')
model.speciesIdentifiers.append('s2')
model.species['s2'] = Species('', 'species 2', '17')
model.reactions.append(
Reaction('r1', 'reaction 1', [SpeciesReference('s1')],
[SpeciesReference('s2')], True, '2+sin(t)'))
model.reactions.append(
Reaction('r2', 'reaction 2',
[SpeciesReference('s1'),
SpeciesReference('s2')], [SpeciesReference('s1', 2)], False,
'1+exp(-t)'))
writeModelXml(model)
print('')
print(model.writeSpeciesTable())
print('')
print(model.writeReactionsTable())
# The model must be evaluated for writing SBML or the C++ code.
model.evaluateInhomogeneous()
writeModelSbml(model, open('model.xml', 'w'))
print('')
print(model.makeInhomogeneousPropensities(True))
print('')
def main(opt):
# Load default values and env
env, envName, nEp, nStep, nRFF, kernelType, lowS, highS, glue, sigma, alphaQ, betaQ, alphaU, betaU = gen_env(
opt)
print(opt)
# Generate directory for experiment
if opt.enableLogFile:
expName, i = "{}_{}".format(opt.expName, envName), 1
while os.path.exists("exps/{}{}".format(expName, i)):
i += 1
dirName = "exps/{}{}".format(expName, i)
os.makedirs(dirName)
# Run experiment
allRets = []
for repI in range(opt.nRepeat):
tm = TransitionMemory(nEp, nStep, glue.dS(), glue.dA())
# Agt, Pol
if opt.featureType == "RFF": # Random Fourier Features
ftmap = RFF(nRFF, glue.dS() + glue.dA(), sigma, kernelType)
elif opt.featureType == "FBF": # Fourier basis features
ftmap = FBF(glue.dS() + glue.dA(), opt.fourierBasisOrder)
agtHlp = AgentHelper(glue, ftmap)
if opt.agent == "method":
agent = Method(agtHlp,
gamma=opt.gamma,
alphaQ=alphaQ,
betaQ=betaQ,
alphaU=alphaU,
betaU=betaU,
tm=tm,
maxVIiter=opt.maxVIiterQ,
kappa=opt.kappa)
elif opt.agent == "QLearning":
lr = ConstantRate(opt.learningRate)
agent = QLearning(agtHlp, gamma=opt.gamma, learningRate=lr)
else:
raise ValueError("Unknown agent {}".format(opt.agent))
policy = eGreedyPolicy(agtHlp, agent.score, opt.epsilon)
# Runner
runner = EnvRunner(glue, policy, agent, tm, opt.verbose)
# Go
runner.run(nEp, nStep, opt.stopWhenSolved)
# Keep track of returns
idx = np.cumsum([0] + tm.getEpisodeLengths())
ss, _, rr, __ = tm.getTransitions()
sMin, sMax = np.min(ss, 0), np.max(ss, 0)
print("State min/max:\n{}\n{}".format(sMin, sMax))
rets = [np.sum(rr[idx[i - 1]:idx[i]]) for i in range(1, len(idx))]
print("Repeat", repI, "finished.")
print("Returns:\n", rets)
allRets.append(rets)
# Parse all variables to file
del rr, _, ss, __
if opt.enableLogFile:
filename = "{}/vars_{}.json".format(dirName, uuid.uuid4().hex)
with open(filename, 'w') as f:
json.dump({k: repr(v)
for k, v in vars().items()},
f,
indent=4,
sort_keys=True)
def parse_methods(self):
""" Parse <method> Nodes """
from Method import Method
for meth in self.find_children(self.ns(self.NS_CORE, 'method')):
self.methods.append(Method(self._namespace, meth))
def ParseMethodName(text):
from settings import PrivateDefinition,PublicDefinition
from settings import ProtectedDefinition,StaticDefinition
from settings import OverrideDefinition,AbstractDefinition
from settings import VirtualDefinition
Parameters=[]
original = text
if (text.find("(")>-1):
Parameters=ParseParameters(text[text.find("(")+1:text.find(")")])
text = text[:text.find("(")]
text=text+" "
original = text
m = Method()
words = text.split()
for word in words:
if (word==PrivateDefinition.replace(" ","")):
m.Private=True
text=text.replace(word,"")
elif (word==PublicDefinition.replace(" ","")):
m.Public =True
text=text.replace(word,"")
elif (word==ProtectedDefinition.replace(" ","")):
m.Protected=True
text=text.replace(word,"")
elif (word==StaticDefinition.replace(" ","")):
m.Static=True
text=text.replace(word,"")
elif (word==OverrideDefinition.replace(" ","")):
m.Override=True
text=text.replace(word,"")
elif (word==AbstractDefinition.replace(" ","")):
m.Abstract=True
text=text.replace(word,"")
elif (word==VirtualDefinition.replace(" ","")):
m.Virtual=True
text=text.replace(word,"")
elif (word in DataTypes):
m.ReturnType=word
text=text.replace(word,"")
if len(text.split())==2:
m.ReturnType=text.split()[0]
m.Name=text.split()[1]
elif len(text.split())==1:
m.ReturnType = ""
m.Name=text.split()[0]
else:
raise ValueError("nerozpoznane paramtre v nazve metody v pocte "+str(len(text.split()))+" text: ",text)
m.Parameters = Parameters
return m
def write(output_file_path,method_instance,tree_function,weather,surface_fuels,simulation,slices):
# function to handle recursive file creation/calling of writer (function that actually does the writing)
# INPUT
# all inputs not given here are the same as writer function - look at writer function documentation below
# tree_function (list or string) - if given as a list file path will be recursively created and output for each element in list
# for what tree_function actually is look at writer documentation
# surface_fuels (list or list of lists) - if a list is given at any index of surface_fuels file path, and output will be recursively
# created for each element. For further documentation look under writer.
# weather (list or list of lists) - if a list is given at any index of surface_fuels file path, and output will be recursively
# created for each element. For further documentation look under writer.
# method_instance (switch case) - can be instance, string, list or a list containing any combination of all three
# (Method instance) - instance of the Method class should hold tree_list with positions
# OR
# (method='csv' info) - [[x,y],inputToMethod.csv] or [[x,y],path\To\Folder\Containing\Csv\Files]
# - [[x,y],inputToMethod.csv] (list) - x,y and .csv file given - Method instance created from input
# OR
# - [[x,y],path\To\Folder\Containing\Csv\Files] (list) - x,y and file path given - Method instances created from every .csv file in file path
# OR
# (method='load' info) - 'Path\To\Folder\Containing\Pkl\Files' or .pkl file
# - 'Path\To\Folder\Containing\Pkl\Files' (str) - path to folder containing .pkl files, Method instance created using load from every .pkl file in path
# - .pkl file (str) - Method instance created using load from .pkl file
# OR
# (list) - list containing any combination of the three above
# output_file_path (string) - path to desired output root file. Files will be created in that folder.
# OUTPUT
# all output created by writer function
# ----note----
# max recursion depth is 1000 by default, if you want to create more than 1000 fds input documents
# you will have to reset max recursion depth (may God have mercy on your soul)
bat_file = open(output_file_path + '\Run.bat', "wb")
def list_me(x):
if type(x) == list:
if len(x) == 2 and (((type(x[0]) == list) or (type(x[0]) == tuple)) and (type(x[1]) == str)):
return [x] # in case method_instance input is just one list of type method='csv' info
else: return x
else: return [x]
method_final = []
for i in list_me(method_instance):
if isinstance(i,Method):
method_final.append(i)
elif type(i) == str:
# file path to folder containing .pkl files
if list(i[1])[-1] == '\\' or list(i[1])[-1] == '/':
i = i + '*.pkl'
else: i = i + '\*.pkl'
files = glob.glob(i)
for j in files:
method_final.append( Method(file_name=j) )
elif type(i) == list:
# list [[x,y], path to .csv file] or [[x,y], path to file containing lots of .csv files]
if ((type(i[0]) == list) or (type(i[0]) == tuple)) and (type(i[1]) == str):
r = re.compile(r'\w+') # regular expression grabs all alpha numeric characters
if r.findall(i[1])[-1] == 'csv': # check to see if it is a csv file
method_final.append( Method(i[0][0],i[0][1],i[1]) )
else:
if list(i[1])[-1] == '\\' or list(i[1])[-1] == '/':
i[1] += '*.csv'
else: i[1] += '\*.csv'
files = glob.glob(i[1])
for j in files:
method_final.append( Method(i[0][0],i[0][1],j) )
else:sys.exit('inputs to write do not match formating convention')
def recursion(output_file_path,method_instance,tree_function,weather,surface_fuels,simulation,slices):
if type(method_instance) == list:
for i in method_instance:
output_file_path += ('\\' + i.file_name)
if not os.path.exists(output_file_path):
os.makedirs(output_file_path)
i.plot_tree_list(output_file_path) # place plot of tree_list in output file
i.plot_vertical_fuel_profile(output_file_path) # place plot of vertical fuel profile in output file
#i.write_full_csv(output_file_path)
recursion(output_file_path,i,tree_function,weather,surface_fuels,simulation,slices)
output_file_path = output_file_path[:-len('\\' + i.file_name)]
elif type(tree_function) == list:
# tree
for i in tree_function:
output_file_path += ('\\' + i.upper())
recursion(output_file_path,method_instance,i,weather,surface_fuels,simulation,slices)
output_file_path = output_file_path[:-len('\\' + i.upper())]
elif type(weather[0]) == list:
# humidity
for i in weather[0]:
output_file_path += ('\Humidity_%s_na' % str(i)) # humidity is a dimensionless ratio na = not applicable
recursion(output_file_path,method_instance,tree_function,[i,weather[1],weather[2]],surface_fuels,simulation,slices)
output_file_path = output_file_path[:-len('\Humidity_%s_na' % str(i))]
elif type(weather[1]) == list:
# wind
for i in weather[1]:
output_file_path += ('\Wind_%s_ms' % str(i)) # ms = m/s
recursion(output_file_path,method_instance,tree_function,[weather[0],i,weather[2]],surface_fuels,simulation,slices)
output_file_path = output_file_path[:-len('\Wind_%s_ms' % str(i))]
elif type(weather[2]) == list:
#init_temp
for i in weather[2]:
output_file_path += ('\Init_temp_%s_dC' % str(i)) # dC = degrees Celsius
recursion(output_file_path,method_instance,tree_function,[weather[0],weather[1],i],surface_fuels,simulation,slices)
output_file_path = output_file_path[:-len('\Init_temp_%s_dC' % str(i))]
elif type(surface_fuels[0]) == list:
#sf_moisture
for i in surface_fuels[0]:
output_file_path += ('\Sf_moisture_%s_na' % str(i)) # moisture content is a dimensionless ratio na = not applicable
recursion(output_file_path,method_instance,tree_function,weather,[i,surface_fuels[1],surface_fuels[2]],simulation,slices)
output_file_path = output_file_path[:-len('\Sf_moisture_%s_na' % str(i))]
elif type(surface_fuels[1]) == list:
#sf_ht
for i in surface_fuels[1]:
output_file_path += ('\Sf_ht_%s_m' % str(i)) # m = meters
recursion(output_file_path,method_instance,tree_function,weather,[surface_fuels[0],i,surface_fuels[2]],simulation,slices)
output_file_path = output_file_path[:-len('\Sf_ht_%s_m' % str(i))]
elif type(surface_fuels[2]) == list:
#sf_load
for i in surface_fuels[2]:
output_file_path += ('\Sf_load_%s_kgM' % str(i)) #kgM = kg per meter squared
recursion(output_file_path,method_instance,tree_function,weather,[surface_fuels[0],surface_fuels[1],i],simulation,slices)
output_file_path = output_file_path[:-len('\Sf_load_%s_kgM' % str(i))]
else:
bat_file.write(output_file_path + '\n')
writer(output_file_path,method_instance,tree_function,weather,surface_fuels,simulation,slices)
recursion(output_file_path,method_final,list_me(tree_function),weather,surface_fuels,simulation,slices)
if __name__ == '__main__':
x = 30
y = 30
tree_function = ['green']#['green','red','mpb_truth']
humidity = .43 # humidity of the air (mass H20/volume air+H20)
wind = 2 #[2,3,4] # wind speed from the "x" wall (m/s)
init_temp = 25 #[20,30] # initial temp of the simulation, this will be the air temp as well as initial veg temp (degrees C)
sf_moisture = 0.3 # surface fuel moisture (weight H20/weight dry)
sf_ht = .5 # [.5,.6] # surface fuel (m)
sf_load = [7.] # surface fuel (kg/m^2)
mesh = 5 # number of meshes in simulation (see note 2 in Write.py) (int)
sim_time = 5 # time of simulation (s)
dt_part = .1 # dump simulation data every dt_part seconds (used in "&DUMP DT_PART" line)(s)
dt_slice = .5 # dump slice every dt_slice seconds (used in "&DUMP DT_SLCF" line) (s)
weather = [humidity,wind,init_temp]
surface_fuels = [sf_moisture,sf_ht,sf_load]
simulation = [mesh,sim_time,dt_part,dt_slice]
#file_name2 = [[x,y],'E:\Projects\STANDFIRE\Data\PLOT1.csv']
#a = Method(x=50,y=50,file_name='E:\Projects\STANDFIRE\Data\PLOT1621331010602.csv',TreePlaceMethod='sr')
a = Method(x=50,y=50,file_name='E:\Projects\STANDFIRE\Data\PLOT1.csv',TreePlaceMethod='sr')
#print a.canopy_bulk_density_rein()
write('E:\Projects\STANDFIRE\Data',a,tree_function,weather,surface_fuels,simulation,None)
def tearDown(self):
Method.tearDown(self)
print('-----------------------------\t')
## method settting
method = 'r0=random0.1'
train_flag = True
train_flag = False
RLmethod = Method(
method,
env.action_dim, # 动作的维度
env.ob_dim, # 状态的维度
env.a_bound, # 动作的上下限
e_greedy_end=0.05, # 最后的探索值 0.1倍幅值
e_liner_times=1500 * 50, # 探索值经历多少次学习变成e_end
epilon_init=0.5, # 表示1倍的幅值作为初始值
LR_A=0.00001, # Actor的学习率
LR_C=0.001, # Critic的学习率
GAMMA=0.9, # 衰减系数
TAU=0.01, # 软替代率，例如0.01表示学习eval网络0.01的值，和原网络0.99的值
MEMORY_SIZE=3000, # 记忆池容量
BATCH_SIZE=128, # 批次数量
units_a=1000, # Actor神经网络单元数
units_c=1000, # Crtic神经网络单元数
actor_learn_start=10000, # Actor开始学习的代数
tensorboard=True, # 是否存储tensorboard
train=train_flag # 训练的时候有探索
)
############################### training ####################################
if RLmethod.train:
for i in range(max_Episodes):
## method settting
method = 'r0=1,4'
# tensorboard --logdir="2 QUAD/6 DL_OP/1 DDPG_OP/logs"
train_flag = True
train_flag = False
RLmethod = Method(
method,
env.lambda_dim, # 动作的维度
env.ob_dim, # 状态的维度
e_greedy_end=0.000, # 最后的探索值 0.1倍幅值
e_liner_times=max_Episodes * 0.2, # 探索值经历多少次学习变成e_end
epilon_init=1, # 表示1倍的幅值作为初始值
LR_A=0.0001, # Actor的学习率
LR_C=0.0001, # Critic的学习率
GAMMA=0.9, # 衰减系数
TAU=0.01, # 软替代率，例如0.01表示学习eval网络0.01的值，和原网络0.99的值
MEMORY_SIZE=20000, # 记忆池容量
BATCH_SIZE=500, # 批次数量
units_a=300, # Actor神经网络单元数
units_c=300, # Crtic神经网络单元数
actor_learn_start=10000, # Actor开始学习的代数
tensorboard=True, # 是否存储tensorboard
train=train_flag # 训练的时候有探索
)
RLmethod.constant = env.constant
############################### training ####################################
if RLmethod.train: