kimbring2

from pysc2.lib import features, point, actions, units
from pysc2.env.environment import TimeStep, StepType
from pysc2.env import sc2_env, available_actions_printer
from pysc2 import run_configs
from s2clientprotocol import sc2api_pb2 as sc_pb

import importlib
import random
import sys
import glob

kimbring2

import tensorflow as tf
import numpy as np


class SelectedUnitsHead(tf.keras.layers.Layer):
  def __init__(self):
    super(SelectedUnitsHead, self).__init__()

    self.model = tf.keras.layers.Dense(256, activation='relu')

kimbring2

import tensorflow as tf
import numpy as np


class ResBlock_MLP(tf.keras.layers.Layer):
    def __init__(self, output_dim, **kwargs):
        self.shortcut = tf.keras.layers.Dense(256, activation='relu')

        self.mlp_0 = tf.keras.layers.Dense(256, activation='relu')
        self.mlp_1 = tf.keras.layers.Dense(256, activation='relu')

kimbring2

from pysc2.env import sc2_env, available_actions_printer
from pysc2.lib import actions, features, units

from utils import get_entity_obs, get_upgrade_obs, get_gameloop_obs, get_race_onehot, get_agent_statistics
from network import Core, ActionTypeHead, SelectedUnitsHead, TargetUnitHead, LocationHead

from absl import flags
FLAGS = flags.FLAGS
FLAGS(sys.argv)

kimbring2

from pysc2.env import sc2_env, available_actions_printer
from pysc2.lib import actions, features, units
import sys
import units_new
import upgrades_new

from utils import get_entity_obs, get_upgrade_obs, get_gameloop_obs, get_race_onehot, get_agent_statistics
from network import EntityEncoder, ScalarEncoder, SpatialEncoder, Core

map_name = 'Simple128'

kimbring2

from pysc2.env import sc2_env, available_actions_printer
from pysc2.lib import actions, features, units

from utils import get_entity_obs, get_upgrade_obs, get_gameloop_obs, get_race_onehot, get_agent_statistics
from network import EntityEncoder, ScalarEncoder, SpatialEncoder, Core

from absl import flags
FLAGS = flags.FLAGS
FLAGS(sys.argv)

kimbring2

import numpy as np

def get_spatial_obs(feature_screen):
  spatial_input  = np.reshape(feature_screen, [1,128,128,27])

  return spatial_input

kimbring2

from pysc2.lib import actions, features, units
import numpy as np
import upgrades_new
import math


def get_agent_statistics(score_by_category):
  score_by_category = score_by_category.flatten()
  agent_statistics = np.log(score_by_category + 1)

kimbring2

from pysc2.lib import actions, features, units
import numpy as np
import units_new
import math

def get_entity_obs(feature_units):
    unit_type = []
    alliance = []
    current_health = []
    current_shields = []

kimbring2

from multiagent import Agent
from pysc2.env import sc2_env, available_actions_printer


def get_supervised_agent(race):
  supervissed_agent = Agent('Terran', None)
  return supervissed_agent


class ActorLoop: