File size: 17,343 Bytes
079c32c |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 |
import argparse
import os
import numpy as np
def str2bool(v):
return v.lower() in ('true', '1')
arg_lists = []
parser = argparse.ArgumentParser()
def add_argument_group(name):
arg = parser.add_argument_group(name)
arg_lists.append(arg)
return arg
# crm
game_arg = add_argument_group('BeerGame')
game_arg.add_argument('--task', type=str, default='bg')
game_arg.add_argument(
'--fixedAction',
type=str2bool,
default='False',
help='if you want to have actions in [0,actionMax] set it to True. with False it will set it [actionLow, actionUp]'
)
game_arg.add_argument(
'--observation_data',
type=str2bool,
default=False,
help='if it is True, then it uses the data that is generated by based on few real world observation'
)
game_arg.add_argument('--data_id', type=int, default=22, help='the default item id for the basket dataset')
game_arg.add_argument('--TLow', type=int, default=100, help='duration of one GAME (lower bound)')
game_arg.add_argument('--TUp', type=int, default=100, help='duration of one GAME (upper bound)')
game_arg.add_argument(
'--demandDistribution',
type=int,
default=0,
help='0=uniform, 1=normal distribution, 2=the sequence of 4,4,4,4,8,..., 3= basket data, 4= forecast data'
)
game_arg.add_argument(
'--scaled', type=str2bool, default=False, help='if true it uses the (if) existing scaled parameters'
)
game_arg.add_argument('--demandSize', type=int, default=6100, help='the size of demand dataset')
game_arg.add_argument('--demandLow', type=int, default=0, help='the lower bound of random demand')
game_arg.add_argument('--demandUp', type=int, default=3, help='the upper bound of random demand')
game_arg.add_argument('--demandMu', type=float, default=10, help='the mu of the normal distribution for demand ')
game_arg.add_argument('--demandSigma', type=float, default=2, help='the sigma of the normal distribution for demand ')
game_arg.add_argument('--actionMax', type=int, default=2, help='it works when fixedAction is True')
game_arg.add_argument(
'--actionUp', type=int, default=2, help='bounds on my decision (upper bound), it works when fixedAction is True'
)
game_arg.add_argument(
'--actionLow', type=int, default=-2, help='bounds on my decision (lower bound), it works when fixedAction is True'
)
game_arg.add_argument(
'--action_step', type=int, default=1, help='The obtained action value by dnn is multiplied by this value'
)
game_arg.add_argument('--actionList', type=list, default=[], help='The list of the available actions')
game_arg.add_argument('--actionListLen', type=int, default=0, help='the length of the action list')
game_arg.add_argument(
'--actionListOpt', type=int, default=0, help='the action list which is used in optimal and sterman'
)
game_arg.add_argument('--actionListLenOpt', type=int, default=0, help='the length of the actionlistopt')
game_arg.add_argument('--agentTypes', type=list, default=['dnn', 'dnn', 'dnn', 'dnn'], help='the player types')
game_arg.add_argument(
'--agent_type1', type=str, default='dnn', help='the player types for agent 1, it can be dnn, Strm, bs, rnd'
)
game_arg.add_argument(
'--agent_type2', type=str, default='dnn', help='the player types for agent 2, it can be dnn, Strm, bs, rnd'
)
game_arg.add_argument(
'--agent_type3', type=str, default='dnn', help='the player types for agent 3, it can be dnn, Strm, bs, rnd'
)
game_arg.add_argument(
'--agent_type4', type=str, default='dnn', help='the player types for agent 4, it can be dnn, Strm, bs, rnd'
)
game_arg.add_argument('--NoAgent', type=int, default=4, help='number of agents, currently it should be in {1,2,3,4}')
game_arg.add_argument('--cp1', type=float, default=2.0, help='shortage cost of player 1')
game_arg.add_argument('--cp2', type=float, default=0.0, help='shortage cost of player 2')
game_arg.add_argument('--cp3', type=float, default=0.0, help='shortage cost of player 3')
game_arg.add_argument('--cp4', type=float, default=0.0, help='shortage cost of player 4')
game_arg.add_argument('--ch1', type=float, default=2.0, help='holding cost of player 1')
game_arg.add_argument('--ch2', type=float, default=2.0, help='holding cost of player 2')
game_arg.add_argument('--ch3', type=float, default=2.0, help='holding cost of player 3')
game_arg.add_argument('--ch4', type=float, default=2.0, help='holding cost of player 4')
game_arg.add_argument('--alpha_b1', type=float, default=-0.5, help='alpha of Sterman formula parameter for player 1')
game_arg.add_argument('--alpha_b2', type=float, default=-0.5, help='alpha of Sterman formula parameter for player 2')
game_arg.add_argument('--alpha_b3', type=float, default=-0.5, help='alpha of Sterman formula parameter for player 3')
game_arg.add_argument('--alpha_b4', type=float, default=-0.5, help='alpha of Sterman formula parameter for player 4')
game_arg.add_argument('--betta_b1', type=float, default=-0.2, help='beta of Sterman formula parameter for player 1')
game_arg.add_argument('--betta_b2', type=float, default=-0.2, help='beta of Sterman formula parameter for player 2')
game_arg.add_argument('--betta_b3', type=float, default=-0.2, help='beta of Sterman formula parameter for player 3')
game_arg.add_argument('--betta_b4', type=float, default=-0.2, help='beta of Sterman formula parameter for player 4')
game_arg.add_argument('--eta', type=list, default=[0, 4, 4, 4], help='the total cost regulazer')
game_arg.add_argument('--distCoeff', type=int, default=20, help='the total cost regulazer')
game_arg.add_argument(
'--ifUseTotalReward',
type=str2bool,
default='False',
help='if you want to have the total rewards in the experience replay, set it to true.'
)
game_arg.add_argument(
'--ifUsedistTotReward',
type=str2bool,
default='True',
help='If use correction to the rewards in the experience replay for all iterations of current game'
)
game_arg.add_argument(
'--ifUseASAO',
type=str2bool,
default='True',
help='if use AS and AO, i.e., received shipment and received orders in the input of DNN'
)
game_arg.add_argument('--ifUseActionInD', type=str2bool, default='False', help='if use action in the input of DNN')
game_arg.add_argument(
'--stateDim', type=int, default=5, help='Number of elements in the state desciptor - Depends on ifUseASAO'
)
game_arg.add_argument('--iftl', type=str2bool, default=False, help='if apply transfer learning')
game_arg.add_argument(
'--ifTransferFromSmallerActionSpace',
type=str2bool,
default=False,
help='if want to transfer knowledge from a network with different action space size.'
)
game_arg.add_argument(
'--baseActionSize',
type=int,
default=5,
help='if ifTransferFromSmallerActionSpace is true, this determines the size of action space of saved network'
)
game_arg.add_argument(
'--tlBaseBrain',
type=int,
default=3,
help='the gameConfig of the base network for re-training with transfer-learning'
)
game_arg.add_argument('--baseDemandDistribution', type=int, default=0, help='same as the demandDistribution')
game_arg.add_argument(
'--MultiAgent', type=str2bool, default=False, help='if run multi-agent RL model, not fully operational'
)
game_arg.add_argument(
'--MultiAgentRun',
type=list,
default=[True, True, True, True],
help='In the multi-RL setting, it determines which agent should get training.'
)
game_arg.add_argument(
'--if_use_AS_t_plus_1', type=str2bool, default='False', help='if use AS[t+1], not AS[t] in the input of DNN'
)
game_arg.add_argument(
'--ifSinglePathExist',
type=str2bool,
default=False,
help='If true it uses the predefined path in pre_model_dir and does not merge it with demandDistribution.'
)
game_arg.add_argument('--gamma', type=float, default=.99, help='discount factor for reward')
game_arg.add_argument(
'--multPerdInpt', type=int, default=10, help='Number of history records which we feed into network'
)
# parameters of the leadtimes
leadtimes_arg = add_argument_group('leadtimes')
leadtimes_arg.add_argument(
'--leadRecItemLow', type=list, default=[2, 2, 2, 4], help='the min lead time for receiving items'
)
leadtimes_arg.add_argument(
'--leadRecItemUp', type=list, default=[2, 2, 2, 4], help='the max lead time for receiving items'
)
leadtimes_arg.add_argument(
'--leadRecOrderLow', type=int, default=[2, 2, 2, 0], help='the min lead time for receiving orders'
)
leadtimes_arg.add_argument(
'--leadRecOrderUp', type=int, default=[2, 2, 2, 0], help='the max lead time for receiving orders'
)
leadtimes_arg.add_argument('--ILInit', type=list, default=[0, 0, 0, 0], help='')
leadtimes_arg.add_argument('--AOInit', type=list, default=[0, 0, 0, 0], help='')
leadtimes_arg.add_argument('--ASInit', type=list, default=[0, 0, 0, 0], help='the initial shipment of each agent')
leadtimes_arg.add_argument('--leadRecItem1', type=int, default=2, help='the min lead time for receiving items')
leadtimes_arg.add_argument('--leadRecItem2', type=int, default=2, help='the min lead time for receiving items')
leadtimes_arg.add_argument('--leadRecItem3', type=int, default=2, help='the min lead time for receiving items')
leadtimes_arg.add_argument('--leadRecItem4', type=int, default=2, help='the min lead time for receiving items')
leadtimes_arg.add_argument('--leadRecOrder1', type=int, default=2, help='the min lead time for receiving order')
leadtimes_arg.add_argument('--leadRecOrder2', type=int, default=2, help='the min lead time for receiving order')
leadtimes_arg.add_argument('--leadRecOrder3', type=int, default=2, help='the min lead time for receiving order')
leadtimes_arg.add_argument('--leadRecOrder4', type=int, default=2, help='the min lead time for receiving order')
leadtimes_arg.add_argument('--ILInit1', type=int, default=0, help='the initial inventory level of the agent')
leadtimes_arg.add_argument('--ILInit2', type=int, default=0, help='the initial inventory level of the agent')
leadtimes_arg.add_argument('--ILInit3', type=int, default=0, help='the initial inventory level of the agent')
leadtimes_arg.add_argument('--ILInit4', type=int, default=0, help='the initial inventory level of the agent')
leadtimes_arg.add_argument('--AOInit1', type=int, default=0, help='the initial arriving order of the agent')
leadtimes_arg.add_argument('--AOInit2', type=int, default=0, help='the initial arriving order of the agent')
leadtimes_arg.add_argument('--AOInit3', type=int, default=0, help='the initial arriving order of the agent')
leadtimes_arg.add_argument('--AOInit4', type=int, default=0, help='the initial arriving order of the agent')
leadtimes_arg.add_argument('--ASInit1', type=int, default=0, help='the initial arriving shipment of the agent')
leadtimes_arg.add_argument('--ASInit2', type=int, default=0, help='the initial arriving shipment of the agent')
leadtimes_arg.add_argument('--ASInit3', type=int, default=0, help='the initial arriving shipment of the agent')
leadtimes_arg.add_argument('--ASInit4', type=int, default=0, help='the initial arriving shipment of the agent')
# test
test_arg = add_argument_group('testing')
test_arg.add_argument(
'--testRepeatMid',
type=int,
default=50,
help='it is number of episodes which is going to be used for testing in the middle of training'
)
test_arg.add_argument('--testInterval', type=int, default=100, help='every xx games compute "test error"')
test_arg.add_argument(
'--ifSaveFigure', type=str2bool, default=True, help='if is it True, save the figures in each testing.'
)
test_arg.add_argument(
'--if_titled_figure',
type=str2bool,
default='True',
help='if is it True, save the figures with details in the title.'
)
test_arg.add_argument(
'--ifsaveHistInterval', type=str2bool, default=False, help='if every xx games save details of the episode'
)
test_arg.add_argument('--saveHistInterval', type=int, default=50000, help='every xx games save details of the play')
test_arg.add_argument('--Ttest', type=int, default=100, help='it defines the number of periods in the test cases')
test_arg.add_argument(
'--ifOptimalSolExist',
type=str2bool,
default=True,
help='if the instance has optimal base stock policy, set it to True, otherwise it should be False.'
)
test_arg.add_argument('--f1', type=float, default=8, help='base stock policy decision of player 1')
test_arg.add_argument('--f2', type=float, default=8, help='base stock policy decision of player 2')
test_arg.add_argument('--f3', type=float, default=0, help='base stock policy decision of player 3')
test_arg.add_argument('--f4', type=float, default=0, help='base stock policy decision of player 4')
test_arg.add_argument(
'--f_init',
type=list,
default=[32, 32, 32, 24],
help='base stock policy decision for 4 time-steps on the C(4,8) demand distribution'
)
test_arg.add_argument('--use_initial_BS', type=str2bool, default=False, help='If use f_init set it to True')
# reporting
reporting_arg = add_argument_group('reporting')
reporting_arg.add_argument('--Rsltdnn', type=list, default=[], help='the result of dnn play tests will be saved here')
reporting_arg.add_argument(
'--RsltRnd', type=list, default=[], help='the result of random play tests will be saved here'
)
reporting_arg.add_argument(
'--RsltStrm', type=list, default=[], help='the result of heuristic fomula play tests will be saved here'
)
reporting_arg.add_argument(
'--Rsltbs', type=list, default=[], help='the result of optimal play tests will be saved here'
)
reporting_arg.add_argument(
'--ifSaveHist',
type=str2bool,
default='False',
help=
'if it is true, saves history, prediction, and the randBatch in each period, WARNING: just make it True in small runs, it saves huge amount of files.'
)
# buildActionList: actions for the beer game problem
def buildActionList(config):
aDiv = 1 # difference in the action list
if config.fixedAction:
actions = list(
range(0, config.actionMax + 1, aDiv)
) # If you put the second argument =11, creates an actionlist from 0..xx
else:
actions = list(range(config.actionLow, config.actionUp + 1, aDiv))
return actions
# specify the dimension of the state of the game
def getStateDim(config):
if config.ifUseASAO:
stateDim = 5
else:
stateDim = 3
if config.ifUseActionInD:
stateDim += 1
return stateDim
def set_optimal(config):
if config.demandDistribution == 0:
if config.cp1 == 2 and config.ch1 == 2 and config.ch2 == 2 and config.ch3 == 2 and config.ch4 == 2:
config.f1 = 8.
config.f2 = 8.
config.f3 = 0.
config.f4 = 0.
def get_config():
config, unparsed = parser.parse_known_args()
config = update_config(config)
return config, unparsed
def fill_leadtime_initial_values(config):
config.leadRecItemLow = [config.leadRecItem1, config.leadRecItem2, config.leadRecItem3, config.leadRecItem4]
config.leadRecItemUp = [config.leadRecItem1, config.leadRecItem2, config.leadRecItem3, config.leadRecItem4]
config.leadRecOrderLow = [config.leadRecOrder1, config.leadRecOrder2, config.leadRecOrder3, config.leadRecOrder4]
config.leadRecOrderUp = [config.leadRecOrder1, config.leadRecOrder2, config.leadRecOrder3, config.leadRecOrder4]
config.ILInit = [config.ILInit1, config.ILInit2, config.ILInit3, config.ILInit4]
config.AOInit = [config.AOInit1, config.AOInit2, config.AOInit3, config.AOInit4]
config.ASInit = [config.ASInit1, config.ASInit2, config.ASInit3, config.ASInit4]
def get_auxuliary_leadtime_initial_values(config):
config.leadRecOrderUp_aux = [config.leadRecOrder1, config.leadRecOrder2, config.leadRecOrder3, config.leadRecOrder4]
config.leadRecItemUp_aux = [config.leadRecItem1, config.leadRecItem2, config.leadRecItem3, config.leadRecItem4]
def fix_lead_time_manufacturer(config):
if config.leadRecOrder4 > 0:
config.leadRecItem4 += config.leadRecOrder4
config.leadRecOrder4 = 0
def set_sterman_parameters(config):
config.alpha_b = [config.alpha_b1, config.alpha_b2, config.alpha_b3, config.alpha_b4]
config.betta_b = [config.betta_b1, config.betta_b2, config.betta_b3, config.betta_b4]
def update_config(config):
config.actionList = buildActionList(config) # The list of the available actions
config.actionListLen = len(config.actionList) # the length of the action list
set_optimal(config)
config.f = [config.f1, config.f2, config.f3, config.f4] # [6.4, 2.88, 2.08, 0.8]
config.actionListLen = len(config.actionList)
if config.demandDistribution == 0:
config.actionListOpt = list(range(0, int(max(config.actionUp * 30 + 1, 3 * sum(config.f))), 1))
else:
config.actionListOpt = list(range(0, int(max(config.actionUp * 30 + 1, 7 * sum(config.f))), 1))
config.actionListLenOpt = len(config.actionListOpt)
config.c_h = [config.ch1, config.ch2, config.ch3, config.ch4]
config.c_p = [config.cp1, config.cp2, config.cp3, config.cp4]
config.stateDim = getStateDim(config) # Number of elements in the state description - Depends on ifUseASAO
get_auxuliary_leadtime_initial_values(config)
fix_lead_time_manufacturer(config)
fill_leadtime_initial_values(config)
set_sterman_parameters(config)
return config
|