-
-
Notifications
You must be signed in to change notification settings - Fork 28
/
Copy pathutils.py
255 lines (204 loc) · 9.61 KB
/
utils.py
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
from pysc2.lib import actions, features, units
import numpy as np
import math
from collections import namedtuple
import os
import random
import collections
import threading
import time
import timeit
from absl import logging
import tensorflow as tf
_PLAYER_RELATIVE = features.SCREEN_FEATURES.player_relative.index
_PLAYER_RELATIVE_SCALE = features.SCREEN_FEATURES.player_relative.scale
_PLAYER_SELF = features.PlayerRelative.SELF
_PLAYER_NEUTRAL = features.PlayerRelative.NEUTRAL
_PLAYER_ENEMY = features.PlayerRelative.ENEMY
_NUM_FUNCTIONS = len(actions.FUNCTIONS)
_SCREEN_PLAYER_ID = features.SCREEN_FEATURES.player_id.index
_SCREEN_UNIT_TYPE = features.SCREEN_FEATURES.unit_type.index
_SCREEN_UNIT_HIT_POINTS = features.SCREEN_FEATURES.unit_hit_points.index
_SCREEN_SELECTED = features.SCREEN_FEATURES.selected.index
_SCREEN_VISIBILITY_MAP = features.SCREEN_FEATURES.visibility_map.index
_MINIMAP_PLAYER_ID = features.MINIMAP_FEATURES.player_id.index
_MINIMAP_CAMERA = features.MINIMAP_FEATURES.camera.index
_MINIMAP_PLAYER_RELATIVE = features.MINIMAP_FEATURES.player_relative.index
all_unit_list = [0, 37, 45, 48, 317, 21, 341, 342, 18, 27, 132, 20, 5, 47, 21,
19, 483, 51, 28, 42, 53, 268, 472, 49, 41, 830, 105, 9, 1680, 110]
# Marine = 48
# Zergling = 105
# Baneling = 9
# Roach = 110
# Mineral = 1680
# Beacon = 317
#essential_unit_list = [0, 45, 48, 317, 21, 341, 18, 27, 20, 19, 483, 500] # For Simple64
#essential_unit_list = [0, 48, 105, 9] # For Minigame
essential_unit_list = [0, 48, 1680]
def preprocess_screen(screen):
layers = []
assert screen.shape[0] == len(features.SCREEN_FEATURES)
for i in range(len(features.SCREEN_FEATURES)):
if i == _SCREEN_UNIT_TYPE:
scale = len(essential_unit_list)
layer = np.zeros([scale, screen.shape[1], screen.shape[2]], dtype=np.float32)
for j in range(len(all_unit_list)):
indy, indx = (screen[i] == all_unit_list[j]).nonzero()
if all_unit_list[j] in essential_unit_list:
unit_index = essential_unit_list.index(all_unit_list[j])
layer[unit_index, indy, indx] = 1
else:
layer[-1, indy, indx] = 1
layers.append(layer)
elif i == _SCREEN_SELECTED:
layer = np.zeros([features.SCREEN_FEATURES[i].scale, screen.shape[1], screen.shape[2]], dtype=np.float32)
for j in range(features.SCREEN_FEATURES[i].scale):
indy, indx = (screen[i] == j).nonzero()
layer[j, indy, indx] = 1
layers.append(layer)
elif i == _SCREEN_UNIT_HIT_POINTS:
layers.append(np.log(screen[i:i+1] + 1) / np.log(features.SCREEN_FEATURES[i].scale))
return np.concatenate(layers, axis=0)
def preprocess_minimap(minimap):
layers = []
assert minimap.shape[0] == len(features.MINIMAP_FEATURES)
for i in range(len(features.MINIMAP_FEATURES)):
if i == features.FeatureType.SCALAR:
layers.append(minimap[i:i+1] / features.MINIMAP_FEATURES[i].scale)
elif i == _MINIMAP_CAMERA or i == _MINIMAP_PLAYER_RELATIVE:
layer = np.zeros([features.MINIMAP_FEATURES[i].scale, minimap.shape[1], minimap.shape[2]], dtype=np.float32)
for j in range(features.MINIMAP_FEATURES[i].scale):
indy, indx = (minimap[i] == j).nonzero()
layer[j, indy, indx] = 1
layers.append(layer)
return np.concatenate(layers, axis=0)
FlatFeature = namedtuple('FlatFeatures', ['index', 'type', 'scale', 'name'])
FLAT_FEATURES = [
FlatFeature(0, features.FeatureType.SCALAR, 1, 'player_id'),
FlatFeature(1, features.FeatureType.SCALAR, 10000, 'minerals'),
FlatFeature(2, features.FeatureType.SCALAR, 10000, 'vespene'),
FlatFeature(3, features.FeatureType.SCALAR, 200, 'food_used'),
FlatFeature(4, features.FeatureType.SCALAR, 200, 'food_cap'),
FlatFeature(5, features.FeatureType.SCALAR, 200, 'food_army'),
FlatFeature(6, features.FeatureType.SCALAR, 200, 'food_workers'),
FlatFeature(7, features.FeatureType.SCALAR, 200, 'idle_worker_count'),
FlatFeature(8, features.FeatureType.SCALAR, 200, 'army_count'),
FlatFeature(9, features.FeatureType.SCALAR, 200, 'warp_gate_count'),
FlatFeature(10, features.FeatureType.SCALAR, 200, 'larva_count'),
]
def preprocess_player(player):
layers = []
for s in FLAT_FEATURES:
if s.index == 1 or s.index == 2:
out = np.log(player[s.index] + 1) / np.log(s.scale)
layers.append(out)
else:
out = player[s.index] / s.scale
layers.append(out)
return np.array(layers)
def preprocess_available_actions(available_action):
available_actions = np.zeros(_NUM_FUNCTIONS, dtype=np.float64)
available_actions[available_action] = 1
return available_actions
def preprocess_feature_units(feature_units, feature_screen_size):
feature_units_list = []
feature_units_length = len(feature_units)
for i, feature_unit in enumerate(feature_units):
#if feature_unit.unit_type == 19:
# print("feature_unit: ", feature_unit)
feature_unit_length = len(feature_unit)
feature_unit_list = []
feature_unit_list.append(feature_unit.unit_type / 2000)
feature_unit_list.append(feature_unit.alliance / 4)
feature_unit_list.append(feature_unit.health / 10000)
feature_unit_list.append(feature_unit.shield / 10000)
feature_unit_list.append(feature_unit.x / 100)
feature_unit_list.append(feature_unit.y / 100)
feature_unit_list.append(feature_unit.is_selected)
feature_unit_list.append(feature_unit.build_progress / 500)
#print("feature_unit.x / (feature_screen_size + 1): ", feature_unit.x / (feature_screen_size + 1))
#print("feature_unit.y / (feature_screen_size + 1): ", feature_unit.y / (feature_screen_size + 1))
feature_units_list.append(feature_unit_list)
#print("i: ", i)
if i >= 49:
break
if feature_units_length < 50:
for i in range(feature_units_length, 50):
feature_units_list.append(np.zeros(8))
entity_array = np.array(feature_units_list)
return entity_array
SingleSelectFeature = namedtuple('SingleSelectFeature', ['index', 'type', 'scale', 'name'])
SINGLE_SELECT_FEATURES = [
SingleSelectFeature(0, features.FeatureType.SCALAR, len(essential_unit_list), 'unit_type'),
SingleSelectFeature(1, features.FeatureType.SCALAR, 4, 'player_relative'),
SingleSelectFeature(2, features.FeatureType.SCALAR, 2000, 'health'),
]
def preprocess_single_select(single_select):
if len(single_select) != 0:
single_select = single_select[0]
layers = []
for s in SINGLE_SELECT_FEATURES:
if s.index == 2:
out = np.log(single_select[s.index] + 1) / np.log(s.scale)
layers.append(out)
elif s.index == 0:
out = essential_unit_list.index(single_select[s.index]) / s.scale
layers.append(out)
else:
out = single_select[s.index] / s.scale
layers.append(out)
else:
layers = [0.0, 0.0, 0.0]
return np.array(layers)
ScoreCumulativeFeature = namedtuple('ScoreCumulativeFeature', ['index', 'type', 'scale', 'name'])
SCORE_CUMULATIVE_FEATURES = [
ScoreCumulativeFeature(0, features.FeatureType.SCALAR, 25000, ' score '),
ScoreCumulativeFeature(1, features.FeatureType.SCALAR, 5000, 'idle_production_time'),
ScoreCumulativeFeature(2, features.FeatureType.SCALAR, 10000, 'idle_worker_time'),
ScoreCumulativeFeature(3, features.FeatureType.SCALAR, 10000, 'total_value_units'),
ScoreCumulativeFeature(4, features.FeatureType.SCALAR, 10000, 'total_value_structures'),
ScoreCumulativeFeature(5, features.FeatureType.SCALAR, 10000, 'killed_value_units'),
ScoreCumulativeFeature(6, features.FeatureType.SCALAR, 10000, 'killed_value_structures'),
ScoreCumulativeFeature(7, features.FeatureType.SCALAR, 10000, 'collected_minerals'),
ScoreCumulativeFeature(8, features.FeatureType.SCALAR, 10000, 'collected_vespene'),
ScoreCumulativeFeature(9, features.FeatureType.SCALAR, 2000, 'collection_rate_minerals'),
ScoreCumulativeFeature(10, features.FeatureType.SCALAR, 2000, 'collection_rate_vespene'),
ScoreCumulativeFeature(11, features.FeatureType.SCALAR, 10000, 'spent_minerals'),
ScoreCumulativeFeature(12, features.FeatureType.SCALAR, 10000, 'spent_vespene'),
]
def preprocess_score_cumulative(score_cumulative):
layers = []
for s in SCORE_CUMULATIVE_FEATURES:
if s.index == 9 or s.index == 10:
out = score_cumulative[s.index] / s.scale
layers.append(out)
else:
out = np.log(score_cumulative[s.index] + 1) / np.log(s.scale)
out = score_cumulative[s.index] / s.scale
layers.append(out)
return np.array(layers)
def preprocess_build_queue(build_queue):
build_queue_length = len(build_queue)
if build_queue_length > 5:
build_queue_length = 5
layers = [0.0, 0.0, 0.0, 0.0, 0.0]
for i in range(0, build_queue_length):
layers[i] = (essential_unit_list.index(build_queue[i][0]) / len(essential_unit_list))
return np.array(layers)
def preprocess_multi_select(multi_select):
multi_select_length = len(multi_select)
if multi_select_length > 10:
multi_select_length = 10
layers = [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
for i in range(0, multi_select_length):
layers[i] = (essential_unit_list.index(multi_select[i][0]) / len(essential_unit_list))
return np.array(layers)
def positional_encoding(max_position, embedding_size, add_batch_dim=False):
positions = np.arange(max_position)
angle_rates = 1 / np.power(10000, (2 * (np.arange(embedding_size)//2)) / np.float32(embedding_size))
angle_rads = positions[:, np.newaxis] * angle_rates[np.newaxis, :]
angle_rads[:, 0::2] = np.sin(angle_rads[:, 0::2])
angle_rads[:, 1::2] = np.cos(angle_rads[:, 1::2])
if add_batch_dim:
angle_rads = angle_rads[np.newaxis, ...]
return tf.cast(angle_rads, dtype=tf.float32)