-
Notifications
You must be signed in to change notification settings - Fork 11
/
model.py
1283 lines (1071 loc) · 59.6 KB
/
model.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
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
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
import torch
import torch.nn as nn
import torch.nn.functional as F
import tools
import math
import numpy as np
from pack_net import LG_RL
class Encoder(nn.Module):
"""Encodes the static & dynamic states using 1d Convolution."""
def __init__(self, input_size, hidden_size):
super(Encoder, self).__init__()
self.conv = nn.Conv1d(input_size, int(hidden_size), kernel_size=1)
def forward(self, input):
output = self.conv(input)
return output # (batch, hidden_size, seq_len)
class HeightmapEncoder(nn.Module):
"""Encodes the static & dynamic states using 1d Convolution."""
def __init__(self, input_size, hidden_size, map_size):
super(HeightmapEncoder, self).__init__()
self.conv1 = nn.Conv2d(input_size, int(hidden_size/4), stride=2, kernel_size=1)
self.conv2 = nn.Conv2d(int(hidden_size/4), int(hidden_size/2), stride=2, kernel_size=1)
self.conv3 = nn.Conv2d(int(hidden_size/2), int(hidden_size), kernel_size=( math.ceil(map_size[0]/4), math.ceil(map_size[1]/4) ) )
def forward(self, input):
output = F.leaky_relu(self.conv1(input))
output = F.leaky_relu(self.conv2(output))
output = self.conv3(output).squeeze(-1)
return output # (batch, hidden_size, seq_len)
class Attention(nn.Module):
"""Calculates attention over the input nodes given the current state."""
def __init__(self, encoder_hidden_size, decoder_hidden_size, decoder_input_type, input_type):
super(Attention, self).__init__()
# W processes features from static decoder elements
self.v = nn.Parameter(torch.zeros((1, 1, decoder_hidden_size), requires_grad=True))
# if input_type == 'use-static' or input_type == 'use-pnet':
# self.W = nn.Parameter(torch.zeros((1, decoder_hidden_size, encoder_hidden_size + decoder_hidden_size), requires_grad=True))
# else:
self.W = nn.Parameter(torch.zeros((1, decoder_hidden_size, 2 * encoder_hidden_size + decoder_hidden_size), requires_grad=True))
self.decoder_input_type = decoder_input_type
self.input_type = input_type
def forward(self, static_hidden, dynamic_hidden, decoder_hidden):
# if self.input_type == 'use-static' or self.input_type == 'use-pnet':
# encoder_hidden = static_hidden
# else:
encoder_hidden = torch.cat( (static_hidden, dynamic_hidden), 1 )
batch_size, hidden_size = decoder_hidden.size()
decoder_hidden = decoder_hidden.unsqueeze(2).repeat(1, 1, static_hidden.shape[-1])
# expand_as(static_hidden)
hidden = torch.cat((encoder_hidden, decoder_hidden), 1)
# Broadcast some dimensions so we can do batch-matrix-multiply
v = self.v.expand(batch_size, 1, hidden_size)
W = self.W.expand(batch_size, hidden_size, -1)
attns = torch.bmm(v, torch.tanh(torch.bmm(W, hidden)))
attns = F.softmax(attns, dim=2) # (batch, seq_len)
return attns
class Pointer(nn.Module):
"""Calculates the next state given the previous state and input embeddings."""
def __init__(self, encoder_hidden_size, decoder_hidden_size, decoder_input_type, input_type, num_layers=1, dropout=0.2):
super(Pointer, self).__init__()
self.encoder_hidden_size = encoder_hidden_size
self.decoder_hidden_size = decoder_hidden_size
self.num_layers = num_layers
self.decoder_input_type = decoder_input_type
self.input_type = input_type
# Used to calculate probability of selecting next state
self.v = nn.Parameter(torch.zeros((1, 1, decoder_hidden_size), requires_grad=True))
# if self.input_type == 'use-static' or self.input_type == 'use-pnet':
# self.W = nn.Parameter(torch.zeros((1, decoder_hidden_size, decoder_hidden_size + encoder_hidden_size), requires_grad=True))
# else:
self.W = nn.Parameter(torch.zeros((1, decoder_hidden_size, 4 * encoder_hidden_size), requires_grad=True))
# Used to compute a representation of the current decoder output
self.gru = nn.GRU( decoder_hidden_size, decoder_hidden_size, num_layers,
batch_first=True,
dropout=dropout if num_layers > 1 else 0)
self.encoder_attn = Attention( encoder_hidden_size, decoder_hidden_size, decoder_input_type, input_type)
self.drop_rnn = nn.Dropout(p=dropout)
self.drop_hh = nn.Dropout(p=dropout)
def forward(self, static_hidden, dynamic_hidden, decoder_hidden, last_hh):
rnn_out, last_hh = self.gru(decoder_hidden.transpose(2, 1), last_hh)
rnn_out = rnn_out.squeeze(1)
# Always apply dropout on the RNN output
rnn_out = self.drop_rnn(rnn_out)
if self.num_layers == 1:
# If > 1 layer dropout is already applied
last_hh = self.drop_hh(last_hh)
# Given a summary of the output, find an input context
enc_attn = self.encoder_attn( static_hidden, dynamic_hidden, rnn_out)
# if self.input_type == 'use-static' or self.input_type == 'use-pnet':
# encoder_hidden = static_hidden
# else:
encoder_hidden = torch.cat( (static_hidden, dynamic_hidden), 1)
context = enc_attn.bmm( encoder_hidden.permute(0, 2, 1)) # (B, 1, num_feats)
# Calculate the next output using Batch-matrix-multiply ops
context = context.transpose(1, 2).expand_as( encoder_hidden )
energy = torch.cat(( encoder_hidden, context), dim=1) # (B, num_feats, seq_len)
v = self.v.expand(static_hidden.size(0), -1, -1)
W = self.W.expand(static_hidden.size(0), -1, -1)
probs = torch.bmm(v, torch.tanh(torch.bmm(W, energy))).squeeze(1)
return probs, last_hh
class DRL(nn.Module):
"""Defines the main Encoder, Decoder, and Pointer combinatorial models.
Parameters
----------
static_size: int
Defines how many features are in the static elements of the model
(e.g. 2 for (x, y) coordinates)
dynamic_size: int > 1
Defines how many features are in the dynamic elements of the model
(e.g. 2 for the VRP which has (load, demand) attributes. The TSP doesn't
have dynamic elements, but to ensure compatility with other optimization
problems, assume we just pass in a vector of zeros.
hidden_size: int
Defines the number of units in the hidden layer for all static, dynamic,
and decoder output units.
update_fn: function or None
If provided, this method is used to calculate how the input dynamic
elements are updated, and is called after each 'point' to the input element.
mask_fn: function or None
Allows us to specify which elements of the input sequence are allowed to
be selected. This is useful for speeding up training of the networks,
by providing a sort of 'rules' guidlines to the algorithm. If no mask
is provided, we terminate the search after a fixed number of iterations
to avoid tours that stretch forever
num_layers: int
Specifies the number of hidden layers to use in the decoder RNN
dropout: float
Defines the dropout rate for the decoder
"""
def __init__(self, static_size, dynamic_size, encoder_hidden_size, decoder_hidden_size,
use_cuda, input_type, allow_rot, container_width, container_height, block_dim,
reward_type, decoder_input_type, heightmap_type, packing_strategy,
update_fn, mask_fn, num_layers=1, dropout=0., unit=1):
super(DRL, self).__init__()
if dynamic_size < 1:
raise ValueError(':param dynamic_size: must be > 0, even if the '
'problem has no dynamic elements')
print(' static size: %d, dynamic size: %d' % (static_size, dynamic_size))
print(' encoder hidden size: %d' % (encoder_hidden_size))
print(' decoder hidden size: %d' % (decoder_hidden_size))
self.update_fn = update_fn
self.mask_fn = mask_fn
# Define the encoder & decoder models
self.static_encoder = Encoder(static_size, encoder_hidden_size)
self.dynamic_encoder = Encoder(dynamic_size, encoder_hidden_size)
heightmap_num = 1
if heightmap_type == 'diff':
if block_dim == 2:
heightmap_width = container_width * unit - 1
elif block_dim == 3:
heightmap_num = 2
heightmap_width = container_width * unit
heightmap_length = container_width * unit
else:
heightmap_width = container_width * unit
if block_dim==3: heightmap_length = container_width * unit
heightmap_width = math.ceil(heightmap_width)
if block_dim==3: heightmap_length = math.ceil(heightmap_length)
if input_type == 'mul' or input_type == 'mul-with':
if block_dim == 2:
heightmap_width = heightmap_width * 2
else:
heightmap_num = heightmap_num * 2
if decoder_input_type == 'shape_only':
self.decoder = Encoder(static_size, decoder_hidden_size)
elif decoder_input_type == 'heightmap_only':
if block_dim == 2:
self.dynamic_decoder = Encoder(heightmap_width, int(decoder_hidden_size))
elif block_dim == 3:
self.dynamic_decoder = HeightmapEncoder(heightmap_num, int(decoder_hidden_size), (heightmap_width, heightmap_length))
elif decoder_input_type == 'shape_heightmap':
self.static_decoder = Encoder(static_size, int(decoder_hidden_size/2))
if block_dim == 2:
self.dynamic_decoder = Encoder(heightmap_width, int(decoder_hidden_size/2))
elif block_dim == 3:
self.dynamic_decoder = HeightmapEncoder(heightmap_num, int(decoder_hidden_size/2), (heightmap_width, heightmap_length))
self.pointer = Pointer(encoder_hidden_size, decoder_hidden_size, decoder_input_type, input_type, num_layers, dropout)
if input_type == 'use-pnet':
self.encoder_RNN = nn.GRU(encoder_hidden_size*2, encoder_hidden_size*2, num_layers,
batch_first=True,
dropout=dropout if num_layers > 1 else 0)
for p in self.parameters():
if len(p.shape) > 1:
nn.init.xavier_uniform_(p)
self.encoder_hidden_size = encoder_hidden_size
self.decoder_hidden_size = decoder_hidden_size
self.use_cuda = use_cuda
self.input_type = input_type
self.allow_rot = allow_rot
self.block_dim = block_dim
self.static_size = static_size
self.dynamic_size = dynamic_size
self.reward_type = reward_type
# if unit < 1:
self.container_width = math.ceil(container_width * unit)
# else:
# self.container_width = container_width * unit
self.container_height = container_height
self.decoder_input_type = decoder_input_type
self.heightmap_type = heightmap_type
self.packing_strategy = packing_strategy
# Used as a proxy initial state in the decoder when not specified
def forward(self, static, dynamic, decoder_input, last_hh=None):
"""
Parameters
----------
static: Array of size (batch_size, feats, num_cities)
Defines the elements to consider as static. For the TSP, this could be
things like the (x, y) coordinates, which won't change
dynamic: Array of size (batch_size, feats, num_cities)
Defines the elements to consider as static. For the VRP, this can be
things like the (load, demand) of each city. If there are no dynamic
elements, this can be set to None
decoder_input: Array of size (batch_size, num_feats)
Defines the outputs for the decoder. Currently, we just use the
static elements (e.g. (x, y) coordinates), but this can technically
be other things as well
last_hh: Array of size (batch_size, num_hidden)
Defines the last hidden state for the RNN
"""
batch_size, _, sequence_size = static.size()
if self.allow_rot == False:
rotate_types = 1
else:
if self.block_dim == 2:
rotate_types = 2
elif self.block_dim == 3:
rotate_types = 6
blocks_num = int(dynamic.shape[-1] / rotate_types)
if self.block_dim == 3:
container_size = [self.container_width, self.container_width, self.container_height]
else:
container_size = [self.container_width, self.container_height]
if self.input_type == 'mul' or self.input_type == 'mul-with':
if self.block_dim == 3:
container_size_a = [self.container_width, self.container_width, self.container_height]
container_size_b = container_size_a
else:
container_size_a = [self.container_width, self.container_height]
container_size_b = container_size_a
if self.input_type == 'mul' or self.input_type == 'mul-with':
containers_a = [tools.Container(container_size_a, blocks_num, self.reward_type, self.heightmap_type, packing_strategy=self.packing_strategy) for _ in range(batch_size)]
containers_b = [tools.Container(container_size_b, blocks_num, self.reward_type, self.heightmap_type, packing_strategy=self.packing_strategy) for _ in range(batch_size)]
else:
containers = [tools.Container(container_size, blocks_num, self.reward_type, self.heightmap_type, packing_strategy=self.packing_strategy) for _ in range(batch_size)]
# Always use a mask - if no function is provided, we don't update it
mask = torch.ones(batch_size, sequence_size)
if self.use_cuda:
mask = mask.cuda()
current_mask = mask.clone()
move_mask = dynamic[:, :blocks_num, :].sum(1)
rotate_small_mask = dynamic[:, blocks_num:blocks_num*2, :].sum(1)
rotate_large_mask = dynamic[:, blocks_num*2:blocks_num*3, :].sum(1)
rotate_mask = rotate_small_mask * rotate_large_mask
dynamic_mask = rotate_mask + move_mask
current_mask[ dynamic_mask.ne(0) ] = 0.
# Structures for holding the output sequences
tour_idx, tour_logp = [], []
max_steps = sequence_size if self.mask_fn is None else 1000
# Static elements only need to be processed once, and can be used across
# all 'pointing' iterations. When / if the dynamic elements change,
# their representations will need to get calculated again.
dynamic_hidden = self.dynamic_encoder(dynamic)
if self.input_type == 'mul':
static_hidden = self.static_encoder(static[:,1:-1,:])
elif self.input_type == 'rot-old':
static_hidden = self.static_encoder(static)
elif self.input_type == 'use-static':
static_hidden = self.static_encoder(static[:,1:,:])
elif self.input_type == 'use-pnet':
static_hidden = self.static_encoder(static[:,1:,:])
# batch_size x dim_num x encoder_hidden_size
static_hidden = static_hidden.transpose(2, 1)
dynamic_hidden = dynamic_hidden.transpose(2, 1)
# RNN for pointer
encoder_hidden, last_hh = self.encoder_RNN( torch.cat( (static_hidden, dynamic_hidden), dim=2 ) )
static_hidden = encoder_hidden[:, :, :self.encoder_hidden_size]
dynamic_hidden = encoder_hidden[:, :, self.encoder_hidden_size:]
static_hidden = static_hidden.transpose(2, 1)
dynamic_hidden = dynamic_hidden.transpose(2, 1)
else:
static_hidden = self.static_encoder(static[:,1:,:])
if 'heightmap' in self.decoder_input_type:
decoder_static, decoder_dynamic = decoder_input
for _ in range(max_steps):
if not mask.byte().any():
break
if self.decoder_input_type == 'shape_only':
decoder_hidden = self.decoder(decoder_input)
elif self.decoder_input_type == 'heightmap_only':
decoder_hidden = self.dynamic_decoder(decoder_dynamic)
elif self.decoder_input_type == 'shape_heightmap':
decoder_static_hidden = self.static_decoder(decoder_static)
decoder_dynamic_hidden = self.dynamic_decoder(decoder_dynamic)
decoder_hidden = torch.cat( (decoder_static_hidden, decoder_dynamic_hidden), 1 )
probs, last_hh = self.pointer(static_hidden,
dynamic_hidden,
decoder_hidden, last_hh)
probs = F.softmax(probs + current_mask.log(), dim=1)
# When training, sample the next step according to its probability.
# During testing, we can take the greedy approach and choose highest
if self.training:
m = torch.distributions.Categorical(probs)
ptr = m.sample()
while not torch.gather(mask, 1, ptr.data.unsqueeze(1)).byte().all():
ptr = m.sample()
logp = m.log_prob(ptr)
else:
prob, ptr = torch.max(probs, 1) # Greedy
logp = prob.log()
# After visiting a node update the dynamic representation
if self.update_fn is not None:
dynamic = self.update_fn(dynamic, static, ptr.data, self.input_type, self.allow_rot)
if self.input_type == 'use-static' or self.input_type == 'use-pnet':
pass
else:
dynamic_hidden = self.dynamic_encoder(dynamic)
# And update the mask so we don't re-visit if we don't need to
if self.mask_fn is not None:
current_mask, mask = self.mask_fn(mask, dynamic, static, ptr.data, self.input_type, self.allow_rot)
current_mask = current_mask.detach()
mask = mask.detach()
if self.input_type == 'mul':
static_part = static[:,1:-1,:]
elif self.input_type == 'rot-old':
static_part = static
else:
static_part = static[:,1:,:]
if self.input_type == 'mul' or self.input_type == 'mul-with':
target_ids = static[:,-1,:]
target_ids = torch.gather( target_ids, 1,
ptr.view(-1, 1)
.expand(-1, 1)).detach()
if 'heightmap' in self.decoder_input_type:
decoder_static = torch.gather( static_part, 2,
ptr.view(-1, 1, 1)
.expand(-1, self.static_size, 1)).detach()
is_rotate = (ptr < blocks_num).cpu().numpy().astype('bool')
if self.input_type == 'mul-with':
blocks = decoder_static.transpose(2,1).squeeze(1).cpu().numpy()[:,:self.block_dim]
else:
blocks = decoder_static.transpose(2,1).squeeze(1).cpu().numpy()
if self.input_type == 'mul' or self.input_type == 'mul-with':
# now get the selected blocks and update heightmap
heightmaps_a = []
heightmaps_b = []
for batch_index in range(batch_size):
target_id = target_ids[batch_index]
if target_id == 0:
heightmaps_a.append( containers_a[batch_index].add_new_block(blocks[batch_index], is_rotate[batch_index] ) )
heightmaps_b.append( containers_b[batch_index].get_heightmap() )
elif target_id == 1:
heightmaps_a.append( containers_a[batch_index].get_heightmap() )
heightmaps_b.append( containers_b[batch_index].add_new_block(blocks[batch_index], is_rotate[batch_index] ) )
if self.block_dim == 2:
if self.use_cuda:
heightmaps_a = torch.FloatTensor(heightmaps_a).cuda().unsqueeze(2)
heightmaps_b = torch.FloatTensor(heightmaps_b).cuda().unsqueeze(2)
else:
heightmaps_a = torch.FloatTensor(heightmaps_a).unsqueeze(2)
heightmaps_b = torch.FloatTensor(heightmaps_b).unsqueeze(2)
decoder_dynamic = torch.cat( (heightmaps_a, heightmaps_b), 1 )
elif self.block_dim == 3:
if self.use_cuda:
heightmaps_a = torch.FloatTensor(heightmaps_a).cuda()#.unsqueeze(1)
heightmaps_b = torch.FloatTensor(heightmaps_b).cuda()#.unsqueeze(1)
else:
heightmaps_a = torch.FloatTensor(heightmaps_a)#.unsqueeze(1)
heightmaps_b = torch.FloatTensor(heightmaps_b)#.unsqueeze(1)
if self.heightmap_type != 'diff':
heightmaps_a = heightmaps_a.unsqueeze(1)
heightmaps_b = heightmaps_b.unsqueeze(1)
decoder_dynamic = torch.cat( (heightmaps_a, heightmaps_b), 1 )
else:
# now get the selected blocks and update heightmap
heightmaps = []
for batch_index in range(batch_size):
heightmaps.append(containers[batch_index].add_new_block(blocks[batch_index], is_rotate[batch_index] ))
if self.block_dim == 2:
if self.use_cuda:
decoder_dynamic = torch.FloatTensor(heightmaps).cuda().unsqueeze(2)
else:
decoder_dynamic = torch.FloatTensor(heightmaps).unsqueeze(2)
elif self.block_dim == 3:
if self.use_cuda:
decoder_dynamic = torch.FloatTensor(heightmaps).cuda()
else:
decoder_dynamic = torch.FloatTensor(heightmaps)
if self.heightmap_type != 'diff':
decoder_dynamic = decoder_dynamic.unsqueeze(1)
else:
decoder_input = torch.gather(static_part, 2,
ptr.view(-1, 1, 1)
.expand(-1, self.static_size, 1)).detach()
# check rotate or not
is_rotate = (ptr < blocks_num).cpu().numpy().astype('bool')
# now get the selected blocks and update containers
if self.input_type == 'mul-with':
blocks = decoder_input.transpose(2,1).squeeze(1).cpu().numpy()[:,:self.block_dim]
else:
blocks = decoder_input.transpose(2,1).squeeze(1).cpu().numpy()
if self.input_type == 'mul' or self.input_type == 'mul-with':
# now get the selected blocks and update heightmap
heightmaps_a = []
heightmaps_b = []
for batch_index in range(batch_size):
target_id = target_ids[batch_index]
if target_id == 0:
containers_a[batch_index].add_new_block(blocks[batch_index], is_rotate[batch_index] )
elif target_id == 1:
containers_b[batch_index].add_new_block(blocks[batch_index], is_rotate[batch_index] )
else:
for batch_index in range(batch_size):
containers[batch_index].add_new_block(blocks[batch_index], is_rotate[batch_index] )
tour_logp.append(logp.unsqueeze(1))
tour_idx.append(ptr.data.unsqueeze(1))
# now we can return the reward at the same time
scores = torch.zeros(batch_size).detach()
if self.use_cuda:
scores = scores.cuda()
if self.input_type == 'mul' or self.input_type == 'mul-with':
for batch_index in range(batch_size):
scores[batch_index] += containers_a[batch_index].calc_ratio()
scores[batch_index] += containers_b[batch_index].calc_ratio()
scores[batch_index] /= 2.0
else:
for batch_index in range(batch_size):
scores[batch_index] = containers[batch_index].calc_ratio()
tour_idx = torch.cat(tour_idx, dim=1) # (batch_size, seq_len)
tour_logp = torch.cat(tour_logp, dim=1) # (batch_size, seq_len)
return tour_idx, tour_logp, None, -scores
class DRL_RNN(nn.Module):
"""Defines the main Encoder, Decoder, and Pointer combinatorial models.
Parameters
----------
static_size: int
Defines how many features are in the static elements of the model
(e.g. 2 for (x, y) coordinates)
dynamic_size: int > 1
Defines how many features are in the dynamic elements of the model
(e.g. 2 for the VRP which has (load, demand) attributes. The TSP doesn't
have dynamic elements, but to ensure compatility with other optimization
problems, assume we just pass in a vector of zeros.
hidden_size: int
Defines the number of units in the hidden layer for all static, dynamic,
and decoder output units.
update_fn: function or None
If provided, this method is used to calculate how the input dynamic
elements are updated, and is called after each 'point' to the input element.
mask_fn: function or None
Allows us to specify which elements of the input sequence are allowed to
be selected. This is useful for speeding up training of the networks,
by providing a sort of 'rules' guidlines to the algorithm. If no mask
is provided, we terminate the search after a fixed number of iterations
to avoid tours that stretch forever
num_layers: int
Specifies the number of hidden layers to use in the decoder RNN
dropout: float
Defines the dropout rate for the decoder
"""
def __init__(self, static_size, dynamic_size, encoder_hidden_size, decoder_hidden_size,
use_cuda, input_type, allow_rot, container_width, container_height, block_dim,
reward_type, decoder_input_type, heightmap_type, packing_strategy,
update_fn, mask_fn, num_layers=1, dropout=0., unit=1):
super(DRL_RNN, self).__init__()
if dynamic_size < 1:
raise ValueError(':param dynamic_size: must be > 0, even if the '
'problem has no dynamic elements')
print(' static size: %d, dynamic size: %d' % (static_size, dynamic_size))
print(' encoder hidden size: %d' % (encoder_hidden_size))
print(' decoder hidden size: %d' % (decoder_hidden_size))
self.update_fn = update_fn
self.mask_fn = mask_fn
# Define the encoder & decoder models
self.static_encoder = Encoder(static_size, encoder_hidden_size)
self.dynamic_encoder = Encoder(dynamic_size, encoder_hidden_size)
heightmap_num = 1
if heightmap_type == 'diff':
if block_dim == 2:
heightmap_width = container_width * unit - 1
elif block_dim == 3:
heightmap_num = 2
heightmap_width = container_width * unit
heightmap_length = container_width * unit
else:
heightmap_width = container_width * unit
if block_dim==3: heightmap_length = container_width * unit
heightmap_width = math.ceil(heightmap_width)
if block_dim==3: heightmap_length = math.ceil(heightmap_length)
if input_type == 'mul' or input_type == 'mul-with':
if block_dim == 2:
heightmap_width = heightmap_width * 2
else:
heightmap_num = heightmap_num * 2
if decoder_input_type == 'shape_only':
self.decoder = Encoder(static_size, decoder_hidden_size)
elif decoder_input_type == 'heightmap_only':
if block_dim == 2:
self.dynamic_decoder = Encoder(heightmap_width, int(decoder_hidden_size))
elif block_dim == 3:
self.dynamic_decoder = HeightmapEncoder(heightmap_num, int(decoder_hidden_size), (heightmap_width, heightmap_length))
elif decoder_input_type == 'shape_heightmap':
self.static_decoder = Encoder(static_size, int(decoder_hidden_size/2))
if block_dim == 2:
self.dynamic_decoder = Encoder(heightmap_width, int(decoder_hidden_size/2))
elif block_dim == 3:
self.dynamic_decoder = HeightmapEncoder(heightmap_num, int(decoder_hidden_size/2), (heightmap_width, heightmap_length))
self.pointer = Pointer(encoder_hidden_size, decoder_hidden_size, decoder_input_type, input_type, num_layers, dropout)
if input_type == 'use-pnet':
self.encoder_RNN = nn.GRU(encoder_hidden_size, encoder_hidden_size, num_layers,
batch_first=True,
dropout=dropout if num_layers > 1 else 0)
for p in self.parameters():
if len(p.shape) > 1:
nn.init.xavier_uniform_(p)
self.encoder_hidden_size = encoder_hidden_size
self.decoder_hidden_size = decoder_hidden_size
self.use_cuda = use_cuda
self.input_type = input_type
self.allow_rot = allow_rot
self.block_dim = block_dim
self.static_size = static_size
self.dynamic_size = dynamic_size
self.reward_type = reward_type
self.container_width = container_width
self.container_height = container_height
self.decoder_input_type = decoder_input_type
self.heightmap_type = heightmap_type
# Used as a proxy initial state in the decoder when not specified
self.packing_strategy = packing_strategy
if self.reward_type == 'C+P+S-G-soft':
self.pack_net = LG_RL.PackRNN(2, 128, container_width, 128, container_width, container_height, heightmap_type, pack_net_type='G')
if packing_strategy[:3] == 'pre':
self.pack_net.load_state_dict(torch.load('./pack_net/G_rand_diff/checkpoints/199/actor.pt'))
print('pre')
if packing_strategy[-4:] == 'eval':
print('eval')
self.pack_net.eval()
else:
print('train')
elif self.reward_type == 'C+P+S-LG-soft':
self.pack_net = LG_RL.PackRNN(2, 128, container_width, 128, container_width, container_height, heightmap_type, pack_net_type='LG')
print('LG')
if packing_strategy[:3] == 'pre':
self.pack_net.load_state_dict(torch.load('./pack_net/LG_rand_diff/checkpoints/199/actor.pt'))
print('pre')
if packing_strategy[-4:] == 'eval':
self.pack_net.eval()
print('eval')
else:
print('train')
else:
print('========> Error in DRL_RNN')
def forward(self, static, dynamic, decoder_input, last_hh=None):
"""
Parameters
----------
static: Array of size (batch_size, feats, num_cities)
Defines the elements to consider as static. For the TSP, this could be
things like the (x, y) coordinates, which won't change
dynamic: Array of size (batch_size, feats, num_cities)
Defines the elements to consider as static. For the VRP, this can be
things like the (load, demand) of each city. If there are no dynamic
elements, this can be set to None
decoder_input: Array of size (batch_size, num_feats)
Defines the outputs for the decoder. Currently, we just use the
static elements (e.g. (x, y) coordinates), but this can technically
be other things as well
last_hh: Array of size (batch_size, num_hidden)
Defines the last hidden state for the RNN
"""
batch_size, _, sequence_size = static.size()
if self.allow_rot == False:
rotate_types = 1
else:
if self.block_dim == 2:
rotate_types = 2
elif self.block_dim == 3:
rotate_types = 6
blocks_num = int(dynamic.shape[-1] / rotate_types)
if self.block_dim == 3:
container_size = [self.container_width, self.container_width, self.container_height]
else:
container_size = [self.container_width, self.container_height]
if self.input_type == 'mul' or self.input_type == 'mul-with':
if self.block_dim == 3:
container_size_a = [self.container_width, self.container_width, self.container_height]
container_size_b = container_size_a
else:
container_size_a = [self.container_width, self.container_height]
container_size_b = container_size_a
# Always use a mask - if no function is provided, we don't update it
mask = torch.ones(batch_size, sequence_size)
if self.use_cuda:
mask = mask.cuda()
current_mask = mask.clone()
move_mask = dynamic[:, :blocks_num, :].sum(1)
rotate_small_mask = dynamic[:, blocks_num:blocks_num*2, :].sum(1)
rotate_large_mask = dynamic[:, blocks_num*2:blocks_num*3, :].sum(1)
rotate_mask = rotate_small_mask * rotate_large_mask
dynamic_mask = rotate_mask + move_mask
current_mask[ dynamic_mask.ne(0) ] = 0.
# Structures for holding the output sequences
tour_idx, tour_logp, pack_logp = [], [], []
max_steps = sequence_size if self.mask_fn is None else 1000
# Static elements only need to be processed once, and can be used across
# all 'pointing' iterations. When / if the dynamic elements change,
# their representations will need to get calculated again.
dynamic_hidden = self.dynamic_encoder(dynamic)
if self.input_type == 'mul':
encoder_static = static[:,1:-1,:]
static_hidden = self.static_encoder(static[:,1:-1,:])
elif self.input_type == 'rot-old':
encoder_static = static
static_hidden = self.static_encoder(static)
elif self.input_type == 'use-static':
encoder_static = static[:, 1:, :]
static_hidden = self.static_encoder(static[:,1:,:])
elif self.input_type == 'use-pnet':
encoder_static = static[:, 1:, :]
static_hidden = self.static_encoder(static[:,1:,:])
# batch_size x dim_num x encoder_hidden_size
static_hidden = static_hidden.transpose(2, 1)
# RNN for pointer
static_hidden, last_hh = self.encoder_RNN(static_hidden)
static_hidden = static_hidden.transpose(2, 1)
else:
encoder_static = static[:, 1:, :]
static_hidden = self.static_encoder(static[:,1:,:])
if 'heightmap' in self.decoder_input_type:
decoder_static, decoder_dynamic = decoder_input
# if self.heightmap_type == 'diff':
# decoder_dynamic = decoder_dynamic[:,:-1, :]
self.P = None
self.C = None
self.S = None
all_blocks = [] * batch_size
# all_rotate = [] * batch_size
# all_order = [] * batch_size
for current_block_num in range(max_steps):
if not mask.byte().any():
break
if self.decoder_input_type == 'shape_only':
decoder_hidden = self.decoder(decoder_input)
elif self.decoder_input_type == 'heightmap_only':
decoder_hidden = self.dynamic_decoder(decoder_dynamic)
elif self.decoder_input_type == 'shape_heightmap':
decoder_static_hidden = self.static_decoder(decoder_static)
decoder_dynamic_hidden = self.dynamic_decoder(decoder_dynamic)
decoder_hidden = torch.cat( (decoder_static_hidden, decoder_dynamic_hidden), 1 )
probs, last_hh = self.pointer(static_hidden,
dynamic_hidden,
decoder_hidden, last_hh)
probs = F.softmax(probs + current_mask.log(), dim=1)
# When training, sample the next step according to its probability.
# During testing, we can take the greedy approach and choose highest
if self.training:
m = torch.distributions.Categorical(probs)
ptr = m.sample()
while not torch.gather(mask, 1, ptr.data.unsqueeze(1)).byte().all():
ptr = m.sample()
logp = m.log_prob(ptr)
else:
prob, ptr = torch.max(probs, 1) # Greedy
logp = prob.log()
# After visiting a node update the dynamic representation
if self.update_fn is not None:
dynamic = self.update_fn(dynamic, static, ptr.data, self.input_type, self.allow_rot)
if self.input_type == 'use-static' or self.input_type == 'use-pnet':
pass
else:
dynamic_hidden = self.dynamic_encoder(dynamic)
# And update the mask so we don't re-visit if we don't need to
if self.mask_fn is not None:
current_mask, mask = self.mask_fn(mask, dynamic, static, ptr.data, self.input_type, self.allow_rot)
current_mask = current_mask.detach()
mask = mask.detach()
if self.input_type == 'mul':
static_part = static[:,1:-1,:]
elif self.input_type == 'rot-old':
static_part = static
else:
static_part = static[:,1:,:]
if self.input_type == 'mul' or self.input_type == 'mul-with':
target_ids = static[:,-1,:]
target_ids = torch.gather( target_ids, 1,
ptr.view(-1, 1)
.expand(-1, 1)).detach()
# if self.reward_type == 'C+P+S-LG-soft' or 'C+P+S-G-soft':
decoder_static = torch.gather( static_part, 2,
ptr.view(-1, 1, 1)
.expand(-1, self.static_size, 1)) #.detach()
# pack current blocks
# TODO multi ?
is_rotate = (ptr < blocks_num).cpu().numpy().astype('bool')
blocks = decoder_static.detach().transpose(2,1).squeeze(1).cpu().numpy()
all_blocks.append(decoder_static)
pack_blocks = torch.cat(all_blocks, dim=-1)
positions, place_logp, scores = self.pack_net( pack_blocks, current_block_num+1 )
# update in container
heightmaps = []
for batch_index in range(batch_size):
heightmaps.append( self.pack_net.engines[batch_index].get_heightap(self.heightmap_type) )
# heightmaps.append(containers[batch_index].add_new_block_at(blocks[batch_index], place_pos[batch_index], is_rotate[batch_index] ))
if self.block_dim == 2:
if self.use_cuda:
decoder_dynamic = torch.FloatTensor(heightmaps).cuda().unsqueeze(2)
else:
decoder_dynamic = torch.FloatTensor(heightmaps).unsqueeze(2)
elif self.block_dim == 3:
if self.use_cuda:
decoder_dynamic = torch.FloatTensor(heightmaps).cuda().unsqueeze(1)
else:
decoder_dynamic = torch.FloatTensor(heightmaps).unsqueeze(1)
tour_logp.append(logp.unsqueeze(1))
tour_idx.append(ptr.data.unsqueeze(1))
tour_idx = torch.cat(tour_idx, dim=1) # (batch_size, seq_len)
tour_logp = torch.cat(tour_logp, dim=1) # (batch_size, seq_len)
# pack_logp = torch.cat(pack_logp, dim=1) # (batch_size, seq_len)
pack_logp = place_logp # (batch_size, seq_len)
# return None
return tour_idx, tour_logp, pack_logp, scores.detach()
class DRL_L(nn.Module):
"""Defines the main Encoder, Decoder, and Pointer combinatorial models.
Parameters
----------
static_size: int
Defines how many features are in the static elements of the model
(e.g. 2 for (x, y) coordinates)
dynamic_size: int > 1
Defines how many features are in the dynamic elements of the model
(e.g. 2 for the VRP which has (load, demand) attributes. The TSP doesn't
have dynamic elements, but to ensure compatility with other optimization
problems, assume we just pass in a vector of zeros.
hidden_size: int
Defines the number of units in the hidden layer for all static, dynamic,
and decoder output units.
update_fn: function or None
If provided, this method is used to calculate how the input dynamic
elements are updated, and is called after each 'point' to the input element.
mask_fn: function or None
Allows us to specify which elements of the input sequence are allowed to
be selected. This is useful for speeding up training of the networks,
by providing a sort of 'rules' guidlines to the algorithm. If no mask
is provided, we terminate the search after a fixed number of iterations
to avoid tours that stretch forever
num_layers: int
Specifies the number of hidden layers to use in the decoder RNN
dropout: float
Defines the dropout rate for the decoder
"""
def __init__(self, static_size, dynamic_size, encoder_hidden_size, decoder_hidden_size,
use_cuda, input_type, allow_rot, container_width, container_height, block_dim,
reward_type, decoder_input_type, heightmap_type, packing_strategy,
update_fn, mask_fn, num_layers=1, dropout=0., unit=1.0):
super(DRL_L, self).__init__()
if dynamic_size < 1:
raise ValueError(':param dynamic_size: must be > 0, even if the '
'problem has no dynamic elements')
print(' static size: %d, dynamic size: %d' % (static_size, dynamic_size))
print(' encoder hidden size: %d' % (encoder_hidden_size))
print(' decoder hidden size: %d' % (decoder_hidden_size))
self.update_fn = update_fn
self.mask_fn = mask_fn
# Define the encoder & decoder models
self.static_encoder = Encoder(static_size, encoder_hidden_size)
self.dynamic_encoder = Encoder(dynamic_size, encoder_hidden_size)
heightmap_num = 1
if heightmap_type == 'diff':
if block_dim == 2:
heightmap_width = container_width * unit - 1
elif block_dim == 3:
heightmap_num = 2
heightmap_width = container_width * unit
heightmap_length = container_width * unit
else:
heightmap_width = container_width * unit
if block_dim==3: heightmap_length = container_width * unit
heightmap_width = math.ceil(heightmap_width)
if block_dim==3: heightmap_length = math.ceil(heightmap_length)
if input_type == 'mul' or input_type == 'mul-with':
if block_dim == 2:
heightmap_width = heightmap_width * 2
else:
heightmap_num = heightmap_num * 2
if decoder_input_type == 'shape_only':
self.decoder = Encoder(static_size, decoder_hidden_size)
elif decoder_input_type == 'heightmap_only':
if block_dim == 2:
self.dynamic_decoder = Encoder(heightmap_width, int(decoder_hidden_size))
elif block_dim == 3:
self.dynamic_decoder = HeightmapEncoder(heightmap_num, int(decoder_hidden_size), (heightmap_width, heightmap_length))
elif decoder_input_type == 'shape_heightmap':
self.static_decoder = Encoder(static_size, int(decoder_hidden_size/2))
if block_dim == 2:
self.dynamic_decoder = Encoder(heightmap_width, int(decoder_hidden_size/2))
elif block_dim == 3:
self.dynamic_decoder = HeightmapEncoder(heightmap_num, int(decoder_hidden_size/2), (heightmap_width, heightmap_length))
self.pointer = Pointer(encoder_hidden_size, decoder_hidden_size, decoder_input_type, input_type, num_layers, dropout)
if input_type == 'use-pnet':
self.encoder_RNN = nn.GRU(encoder_hidden_size, encoder_hidden_size, num_layers,
batch_first=True,
dropout=dropout if num_layers > 1 else 0)
for p in self.parameters():
if len(p.shape) > 1:
nn.init.xavier_uniform_(p)
self.encoder_hidden_size = encoder_hidden_size
self.decoder_hidden_size = decoder_hidden_size
self.use_cuda = use_cuda
self.input_type = input_type
self.allow_rot = allow_rot
self.block_dim = block_dim
self.static_size = static_size
self.dynamic_size = dynamic_size
self.reward_type = reward_type
self.container_width = container_width
self.container_height = container_height
self.heightmap_type = heightmap_type
self.decoder_input_type = decoder_input_type
self.packing_strategy = packing_strategy
if reward_type == 'C+P+S-RL-soft':
print('RL')
if heightmap_type == 'diff':
self.pack_net = tools.DQN(self.container_width, is_diff_height=True)
if packing_strategy[:3] == 'pre':
self.pack_net.load_state_dict(torch.load('./pack_net/RL_rand_diff/checkpoints/199/actor.pt'))
if packing_strategy[-4:] == 'eval':
print('eval')
self.pack_net.eval()
elif heightmap_type == 'normal':
self.pack_net = tools.DQN(self.container_width, is_diff_height=False)
if packing_strategy[:3] == 'pre':
self.pack_net.load_state_dict(torch.load('./pack_net/RL_rand_normal/checkpoints/199/actor.pt'))
if packing_strategy[-4:] == 'eval':