-
Notifications
You must be signed in to change notification settings - Fork 0
/
model.py
806 lines (718 loc) · 36.2 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
import math
import torch
import torch.nn as nn
from torch.autograd import Variable
import torch.nn.functional as F
from torch.nn.utils.rnn import pack_padded_sequence, pad_packed_sequence
import numpy as np
from utils import try_cuda
from attn import MultiHeadAttention
from env_simplified import ConvolutionalImageFeatures, BottomUpImageFeatures
from KB.ConceptNet import GCN_pre_define
def make_image_attention_layers(args, image_features_list, hidden_size):
image_attention_size = args.image_attention_size or hidden_size
attention_mechs = []
for featurizer in image_features_list:
if isinstance(featurizer, ConvolutionalImageFeatures):
if args.image_attention_type == 'feedforward':
attention_mechs.append(MultiplicativeImageAttention(
hidden_size, image_attention_size,
image_feature_size=featurizer.feature_dim))
elif args.image_attention_type == 'multiplicative':
attention_mechs.append(FeedforwardImageAttention(
hidden_size, image_attention_size,
image_feature_size=featurizer.feature_dim))
elif isinstance(featurizer, BottomUpImageFeatures):
attention_mechs.append(BottomUpImageAttention(
hidden_size,
args.bottom_up_detection_embedding_size,
args.bottom_up_detection_embedding_size,
image_attention_size,
featurizer.num_objects,
featurizer.num_attributes,
featurizer.feature_dim
))
else:
attention_mechs.append(None)
attention_mechs = [
try_cuda(mech) if mech else mech for mech in attention_mechs]
return attention_mechs
# TODO: make all attention module return logit instead of weight
class PositionalEncoding(nn.Module):
def __init__(self, d_model, dropout, max_len=80):
super(PositionalEncoding, self).__init__()
self.dropout = nn.Dropout(p=dropout)
# Compute the PE once
pe = torch.zeros(max_len, d_model)
position = torch.arange(0, max_len).unsqueeze(1).float()
div_term = torch.exp(
torch.arange(0, d_model, 2).float() / d_model * \
(-math.log(10000.0)))
pe[:,0::2] = torch.sin(position * div_term)
pe[:,1::2] = torch.cos(position * div_term)
pe = pe.unsqueeze(0)
self.register_buffer('pe',pe)
def forward(self, x):
x = x + Variable(self.pe[:,:x.size(1)], requires_grad=False)
return self.dropout(x)
# TODO: try variational dropout (or zoneout?)
class EncoderLSTM(nn.Module):
''' Encodes navigation instructions, returning hidden state context (for
attention methods) and a decoder initial state. '''
def __init__(self, vocab_size, embedding_size, hidden_size, padding_idx,
dropout_ratio, bidirectional=False, num_layers=1, glove=None):
super(EncoderLSTM, self).__init__()
self.embedding_size = embedding_size
self.hidden_size = hidden_size
self.drop = nn.Dropout(p=dropout_ratio)
self.num_directions = 2 if bidirectional else 1
self.num_layers = num_layers
self.embedding = nn.Embedding(vocab_size, embedding_size, padding_idx)
self.use_glove = glove is not None
if self.use_glove:
print('Using GloVe embedding')
self.embedding.weight.data[...] = torch.from_numpy(glove)
self.embedding.weight.requires_grad = False
self.lstm = nn.LSTM(embedding_size, hidden_size, self.num_layers,
batch_first=True, dropout=(dropout_ratio if self.num_layers > 1 else 0),
bidirectional=bidirectional)
self.encoder2decoder = nn.Linear(hidden_size * self.num_directions,
hidden_size * self.num_directions)
def init_state(self, batch_size):
''' Initialize to zero cell states and hidden states.'''
h0 = Variable(torch.zeros(
self.num_layers * self.num_directions,
batch_size,
self.hidden_size
), requires_grad=False)
c0 = Variable(torch.zeros(
self.num_layers * self.num_directions,
batch_size,
self.hidden_size
), requires_grad=False)
return try_cuda(h0), try_cuda(c0)
def forward(self, inputs, lengths, seq_mask):
''' Expects input vocab indices as (batch, seq_len). Also requires a
list of lengths for dynamic batching. '''
batch_size = inputs.size(0)
embeds = self.embedding(inputs) # (batch, seq_len, embedding_size)
if not self.use_glove:
embeds = self.drop(embeds)
h0, c0 = self.init_state(batch_size)
packed_embeds = pack_padded_sequence(embeds, lengths, batch_first=True)
enc_h, (enc_h_t, enc_c_t) = self.lstm(packed_embeds, (h0, c0))
if self.num_directions == 2:
h_t = torch.cat((enc_h_t[-1], enc_h_t[-2]), 1)
c_t = torch.cat((enc_c_t[-1], enc_c_t[-2]), 1)
else:
h_t = enc_h_t[-1]
c_t = enc_c_t[-1] # (batch, hidden_size)
decoder_init = nn.Tanh()(self.encoder2decoder(h_t))
ctx, lengths = pad_packed_sequence(enc_h, batch_first=True)
ctx = self.drop(ctx)
# (batch, seq_len, hidden_size*num_directions), (batch, hidden_size)
return ctx, decoder_init, c_t
class SoftDotMultiHead(nn.Module):
'''Soft Dot Attention.
Ref: http://www.aclweb.org/anthology/D15-1166
Adapted from PyTorch OPEN NMT.
'''
def __init__(self, dim, num_head):
'''Initialize layer.'''
super(SoftDotMultiHead, self).__init__()
self.multi = MultiHeadAttention(num_head, dim, dim, dim)
def forward(self, h, k, v, mask=None):
'''Propagate h through the network.
h: batch x dim
k,v: batch x seq_len x dim
mask: batch x seq_len indices to be masked
'''
output, attn = self.multi(h.unsqueeze(1), k, v, mask.unsqueeze(1))
return output.squeeze(1), attn.squeeze(1)
class SoftDotAttention(nn.Module):
'''Soft Dot Attention.
Ref: http://www.aclweb.org/anthology/D15-1166
Adapted from PyTorch OPEN NMT.
'''
def __init__(self, dim):
'''Initialize layer.'''
super(SoftDotAttention, self).__init__()
self.dim = dim
self.linear_in = nn.Linear(dim, dim, bias=False)
self.sm = nn.Softmax(dim=1)
self.linear_out = nn.Linear(dim * 2, dim, bias=False)
self.tanh = nn.Tanh()
def forward(self, h, context, mask=None):
'''Propagate h through the network.
h: batch x dim
context: batch x seq_len x dim
mask: batch x seq_len indices to be masked
'''
target = self.linear_in(h).unsqueeze(2) # batch x dim x 1
# Get attention
attn = torch.bmm(context, target).squeeze(2) # batch x seq_len
# TODO: attn = attn / math.sqrt(self.dim) # prevent extreme softmax
if mask is not None:
# -Inf masking prior to the softmax
attn.data.masked_fill_(mask, -float('inf'))
attn = self.sm(attn)
attn3 = attn.view(attn.size(0), 1, attn.size(1)) # batch x 1 x seq_len
weighted_context = torch.bmm(attn3, context).squeeze(1) # batch x dim
h_tilde = torch.cat((weighted_context, h), 1)
h_tilde = self.tanh(self.linear_out(h_tilde))
return h_tilde, attn
class ContextOnlySoftDotAttention(nn.Module):
def __init__(self, dim, context_dim=None):
'''Initialize layer.'''
super(ContextOnlySoftDotAttention, self).__init__()
if context_dim is None:
context_dim = dim
self.linear_in = nn.Linear(dim, context_dim, bias=False)
self.sm = nn.Softmax(dim=1)
def forward(self, h, context, mask=None):
'''Propagate h through the network.
h: batch x dim
context: batch x seq_len x dim
mask: batch x seq_len indices to be masked
'''
target = self.linear_in(h).unsqueeze(2) # batch x dim x 1
# Get attention
attn = torch.bmm(context, target).squeeze(2) # batch x seq_len
if mask is not None:
# -Inf masking prior to the softmax
attn.data.masked_fill_(mask, -float('inf'))
attn = self.sm(attn)
attn3 = attn.view(attn.size(0), 1, attn.size(1)) # batch x 1 x seq_len
weighted_context = torch.bmm(attn3, context).squeeze(1) # batch x dim
return weighted_context, attn
class FeedforwardImageAttention(nn.Module):
def __init__(self, context_size, hidden_size, image_feature_size=2048):
super(FeedforwardImageAttention, self).__init__()
self.feature_size = image_feature_size
self.context_size = context_size
self.hidden_size = hidden_size
self.fc1_feature = nn.Conv2d(
image_feature_size, hidden_size, kernel_size=1, bias=False)
self.fc1_context = nn.Linear(context_size, hidden_size, bias=True)
self.fc2 = nn.Conv2d(hidden_size, 1, kernel_size=1, bias=True)
def forward(self, feature, context):
batch_size = feature.shape[0]
feature_hidden = self.fc1_feature(feature)
context_hidden = self.fc1_context(context)
context_hidden = context_hidden.unsqueeze(-1).unsqueeze(-1)
x = feature_hidden + context_hidden
x = self.fc2(F.relu(x))
# batch_size x (width * height) x 1
attention = F.softmax(x.view(batch_size, -1), 1).unsqueeze(-1)
# batch_size x feature_size x (width * height)
reshaped_features = feature.view(batch_size, self.feature_size, -1)
x = torch.bmm(reshaped_features, attention) # batch_size x
return x.squeeze(-1), attention.squeeze(-1)
class MultiplicativeImageAttention(nn.Module):
def __init__(self, context_size, hidden_size, image_feature_size=2048):
super(MultiplicativeImageAttention, self).__init__()
self.feature_size = image_feature_size
self.context_size = context_size
self.hidden_size = hidden_size
self.fc1_feature = nn.Conv2d(
image_feature_size, hidden_size, kernel_size=1, bias=True)
self.fc1_context = nn.Linear(context_size, hidden_size, bias=True)
self.fc2 = nn.Conv2d(hidden_size, 1, kernel_size=1, bias=True)
def forward(self, feature, context):
batch_size = feature.shape[0]
# batch_size x hidden_size x width x height
feature_hidden = self.fc1_feature(feature)
# batch_size x hidden_size
context_hidden = self.fc1_context(context)
# batch_size x 1 x hidden_size
context_hidden = context_hidden.unsqueeze(-2)
# batch_size x hidden_size x (width * height)
feature_hidden = feature_hidden.view(batch_size, self.hidden_size, -1)
# batch_size x 1 x (width x height)
x = torch.bmm(context_hidden, feature_hidden)
# batch_size x (width * height) x 1
attention = F.softmax(x.view(batch_size, -1), 1).unsqueeze(-1)
# batch_size x feature_size x (width * height)
reshaped_features = feature.view(batch_size, self.feature_size, -1)
x = torch.bmm(reshaped_features, attention) # batch_size x
return x.squeeze(-1), attention.squeeze(-1)
class BottomUpImageAttention(nn.Module):
def __init__(self, context_size, object_embedding_size,
attribute_embedding_size, hidden_size, num_objects,
num_attributes, image_feature_size=2048):
super(BottomUpImageAttention, self).__init__()
self.context_size = context_size
self.object_embedding_size = object_embedding_size
self.attribute_embedding_size = attribute_embedding_size
self.hidden_size = hidden_size
self.num_objects = num_objects
self.num_attributes = num_attributes
self.feature_size = (image_feature_size + object_embedding_size +
attribute_embedding_size + 1 + 5)
self.object_embedding = nn.Embedding(
num_objects, object_embedding_size)
self.attribute_embedding = nn.Embedding(
num_attributes, attribute_embedding_size)
self.fc1_context = nn.Linear(context_size, hidden_size)
self.fc1_feature = nn.Linear(self.feature_size, hidden_size)
# self.fc1 = nn.Linear(context_size + self.feature_size, hidden_size)
self.fc2 = nn.Linear(hidden_size, 1)
def forward(self, bottom_up_features, context):
# image_features: batch_size x max_num_detections x feature_size
# object_ids: batch_size x max_num_detections
# attribute_ids: batch_size x max_num_detections
# no_object_mask: batch_size x max_num_detections
# context: batch_size x context_size
# batch_size x max_num_detections x embedding_size
attribute_embedding = self.attribute_embedding(
bottom_up_features.attribute_indices)
# batch_size x max_num_detections x embedding_size
object_embedding = self.object_embedding(
bottom_up_features.object_indices)
# batch_size x max_num_detections x (feat size)
feats = torch.cat((
bottom_up_features.cls_prob.unsqueeze(2),
bottom_up_features.image_features,
attribute_embedding, object_embedding,
bottom_up_features.spatial_features), dim=2)
# attended_feats = feats.mean(dim=1)
# attention = None
# batch_size x 1 x hidden_size
x_context = self.fc1_context(context).unsqueeze(1)
# batch_size x max_num_detections x hidden_size
x_feature = self.fc1_feature(feats)
# batch_size x max_num_detections x hidden_size
x = x_context * x_feature
x = x / torch.norm(x, p=2, dim=2, keepdim=True)
x = self.fc2(x).squeeze(-1) # batch_size x max_num_detections
x.data.masked_fill_(bottom_up_features.no_object_mask, -float("inf"))
# batch_size x 1 x max_num_detections
attention = F.softmax(x, 1).unsqueeze(1)
# batch_size x feat_size
attended_feats = torch.bmm(attention, feats).squeeze(1)
return attended_feats, attention
class WhSoftDotAttentionCompact(nn.Module):
''' Visual Dot Attention Layer. '''
def __init__(self, dim, context_dim):
'''Initialize layer.'''
super(WhSoftDotAttentionCompact, self).__init__()
if dim != context_dim:
dot_dim = min(dim, context_dim)
self.linear_in = nn.Linear(dim, dot_dim//2, bias=True)
self.linear_in_2 = nn.Linear(context_dim, dot_dim//2, bias=True)
self.dim = dim
self.context_dim = context_dim
self.sm = nn.Softmax(dim=1)
def forward(self, h, ctx, mask=None, v=None):
if self.dim != self.context_dim:
target = self.linear_in(h).unsqueeze(2) # batch x dim x 1
context = self.linear_in_2(ctx)
else:
target = h.unsqueeze(2)
context = ctx
attn = torch.bmm(context, target).squeeze(2) # batch x seq_len
if mask is not None:
# -Inf masking prior to the softmax
attn.data.masked_fill_(mask, -float('inf'))
attn_sm = self.sm(attn)
attn3 = attn_sm.view(attn.size(0), 1, attn.size(1)) # batch x 1 x seq_len
context = v if v is not None else ctx
weighted_context = torch.bmm(attn3, context).squeeze(1) # batch x dim
return weighted_context, attn
class VisualSoftDotAttention(nn.Module):
''' Visual Dot Attention Layer. '''
def __init__(self, h_dim, v_dim, dot_dim=256):
'''Initialize layer.'''
super(VisualSoftDotAttention, self).__init__()
self.linear_in_h = nn.Linear(h_dim, dot_dim, bias=True)
self.linear_in_v = nn.Linear(v_dim, dot_dim, bias=True)
self.sm = nn.Softmax(dim=1)
def forward(self, h, k, mask=None, v=None):
'''Propagate h through the network.
h: batch x h_dim
k: batch x v_num x v_dim
'''
target = self.linear_in_h(h).unsqueeze(2) # batch x dot_dim x 1
context = self.linear_in_v(k) # batch x v_num x dot_dim
attn = torch.bmm(context, target).squeeze(2) # batch x v_num
attn_sm = self.sm(attn)
attn3 = attn_sm.view(attn.size(0), 1, attn.size(1)) # batch x 1 x v_num
ctx = v if v is not None else k
weighted_context = torch.bmm(
attn3, ctx).squeeze(1) # batch x v_dim
return weighted_context, attn
class WhSoftDotAttention(nn.Module):
''' Visual Dot Attention Layer. '''
def __init__(self, h_dim, v_dim=None):
'''Initialize layer.'''
super(WhSoftDotAttention, self).__init__()
if v_dim is None:
v_dim = h_dim
self.h_dim = h_dim
self.v_dim = v_dim
self.linear_in_h = nn.Linear(h_dim, v_dim, bias=True)
self.sm = nn.Softmax(dim=1)
def forward(self, h, k, mask=None, v=None):
'''Propagate h through the network.
h: batch x h_dim
k: batch x v_num x v_dim
'''
target = self.linear_in_h(h).unsqueeze(2) # batch x dot_dim x 1
attn = torch.bmm(k, target).squeeze(2) # batch x v_num
#attn /= math.sqrt(self.v_dim) # scaled dot product attention
if mask is not None:
attn.data.masked_fill_(mask, -float('inf'))
attn_sm = self.sm(attn)
attn3 = attn_sm.view(attn.size(0), 1, attn.size(1)) # batch x 1 x v_num
ctx = v if v is not None else k
weighted_context = torch.bmm(
attn3, ctx).squeeze(1) # batch x v_dim
return weighted_context, attn
###############################################################################
# transformer models
###############################################################################
class EmbeddingEncoder(nn.Module):
def __init__(self, vocab_size, embedding_size, hidden_size, padding_idx,
dropout_ratio, bidirectional=False, num_layers=1, glove=None):
super(EmbeddingEncoder, self).__init__()
self.embedding_size = embedding_size
self.hidden_size = hidden_size
self.num_layers = num_layers
self.num_directions = 2 if bidirectional else 1
self.embedding = nn.Embedding(vocab_size, embedding_size, padding_idx)
self.position_encoding = PositionalEncoding(hidden_size, dropout_ratio)
self.use_glove = glove is not None
self.fc = nn.Linear(embedding_size, hidden_size)
nn.init.xavier_normal_(self.fc.weight)
if self.use_glove:
print('Using GloVe embedding')
self.embedding.weight.data[...] = torch.from_numpy(glove)
self.embedding.weight.requires_grad = False
def init_state(self, batch_size):
''' Initialize to zero cell states and hidden states.'''
h0 = torch.zeros(batch_size,
self.hidden_size*self.num_layers*self.num_directions,
requires_grad=False)
c0 = torch.zeros(batch_size,
self.hidden_size*self.num_layers*self.num_directions,
requires_grad=False)
return try_cuda(h0), try_cuda(c0)
def forward(self, inputs, lengths,seq_mask=None,nomap=False):
batch_size = inputs.size(0)
embeds = self.embedding(inputs)
if nomap:
return (embeds, *self.init_state(batch_size))
embeds = self.fc(embeds)
embeds = self.position_encoding(embeds)
max_len = max(lengths)
embeds = embeds[:,:max_len,:]
return (embeds, *self.init_state(batch_size))
class TransformerEncoder(nn.Module):
def __init__(self, vocab_size, embedding_size, hidden_size, padding_idx,
dropout_ratio, bidirectional=False, num_layers=2,nhead=6, glove=None,ff=2048):
super(TransformerEncoder, self).__init__()
self.embedding = nn.Embedding(vocab_size, embedding_size, padding_idx)
self.use_glove = glove is not None
if self.use_glove:
print('Using GloVe embedding')
self.embedding.weight.data[...] = torch.from_numpy(glove)
self.embedding.weight.requires_grad = False
encoder_layer = torch.nn.TransformerEncoderLayer(d_model=embedding_size,nhead=nhead,dropout=dropout_ratio,dim_feedforward=ff)
self.transformer_encoder = nn.TransformerEncoder(encoder_layer, num_layers=num_layers)
self.position_embedding = PositionalEncoding(embedding_size,dropout=0)
self.fc = nn.Linear(embedding_size,hidden_size)
def forward(self,inputs,lengths,seq_mask):
batch_size = inputs.size(0)
embeds = self.embedding(inputs) # (batch, seq_len, embedding_size)
if not self.use_glove:
embeds = self.drop(embeds)
seq_mask = seq_mask.bool()
embeds = self.position_embedding(embeds)
max_len = max(lengths)
embeds = embeds[:,:max_len,:]
embeds = embeds.transpose(0,1)
output = self.transformer_encoder(embeds,src_key_padding_mask=seq_mask)
output = output.transpose(0,1)
output = self.fc(output)
lengths = try_cuda(torch.tensor(lengths))
h_0 = output[torch.arange(batch_size),lengths-1,:]
return(output,h_0,h_0)
def hard_softmax(y,dim=1):
shape = y.size()
_, ind = y.max(dim=-1)
y_hard = torch.zeros_like(y).view(-1, shape[-1])
y_hard.scatter_(1, ind.view(-1, 1), 1)
y_hard = y_hard.view(*shape)
# Set gradients w.r.t. y_hard gradients w.r.t. y
y_hard = (y_hard - y).detach() + y
return y_hard
class object_roomTransformer(nn.Module):
#fake code:
#1.language_input->attn_with h_0 get text_object_attn(1*300),text_room_attn(1*512)
#2.text_object_attn(1*300) -> attn_weight with knowledge base get top K(5*300)
# ->gather with top K 1*512()-> refine_object_attn
#3.text_room_attn -> text_room_label ->room_label_text
# ->text_room_loss
#4.label_set -> viewpoint_label -> GCN_with_predefine edges
# ->label_feature_list->atten with refine_obejct_attn
#5.label_set ->object_top_5 ->room_label-> view_room_loss
#find label for room
def __init__(self, embedding_size, hidden_size, dropout_ratio,
feature_size=2048+128, image_attention_layers=None,
visual_hidden_size=1024,num_head=8,num_layer=6,concate_room=False,
wo_instr_input=False,transformer_dropout_rate=0.1,action_prediction_mode='single',label_size=300,
use_room=True,use_object=True,num_gcn=3,gcn_type='in',
max_degree=10,short_cut=False,ff=2048,label_length = 5,soft_room_label=True,room_relation_vec=True,object_top_n=5,load_room_relation_weight=True,load_object_relation_weight=True):
super(object_roomTransformer,self).__init__()
self.embedding_size = embedding_size
self.feature_size = feature_size
self.action_prediction_mode = action_prediction_mode
self.wo_instr_input = wo_instr_input
self.hidden_size = hidden_size
self.search = False
self.topk = object_top_n
self.u_begin = try_cuda(Variable(
torch.zeros(embedding_size), requires_grad=False))
self.history_begin = try_cuda(Variable(
torch.zeros(1,hidden_size), requires_grad=False)
)
self.positional_encoding = PositionalEncoding(hidden_size, dropout=0)
self.drop = nn.Dropout(p=dropout_ratio)
self.use_room = use_room
self.use_object = use_object
self.soft_room_label = soft_room_label
num_rooms = 31
repeat = 128
decoder_layer =nn.TransformerDecoderLayer(hidden_size,num_head,dropout=transformer_dropout_rate,dim_feedforward=ff)
self.decoder = nn.TransformerDecoder(decoder_layer,num_layer)
self.object_text_attention_layer = WhSoftDotAttention(hidden_size, hidden_size)
self.room_text_attention_layer = WhSoftDotAttention(hidden_size,hidden_size)
self.object_map = nn.Linear(hidden_size,300)
self.text_room_classifier = nn.Sequential(
nn.Linear(hidden_size, hidden_size),
nn.Dropout(dropout_ratio),
nn.Linear(hidden_size,31),
nn.Softmax(dim=1)
)
for p in self.text_room_classifier.parameters():
p.requires_grad=False
self.view_room_classifier = nn.Sequential(
nn.Linear(label_length*300, hidden_size),
nn.Dropout(dropout_ratio),
nn.Linear(hidden_size,31),
nn.Softmax(dim=1)
)
for p in self.view_room_classifier.parameters():
p.requires_grad=False
self.sm = nn.Softmax(dim=1)
self.object_attention_layer = WhSoftDotAttention(300,300)
self.object_refine_layer = WhSoftDotAttention(300,300)
self.room_attention_layer = WhSoftDotAttention(hidden_size,300)
self.room_relation_vec = room_relation_vec
if room_relation_vec:
self.room_class_relation = torch.nn.Parameter(torch.randn(num_rooms,num_rooms,repeat),requires_grad=True)
else:
if load_room_relation_weight:
x = np.load('KB/data/relations_room.npy')
self.room_class_relation = torch.nn.Parameter(torch.relu(torch.from_numpy(x).float()),requires_grad=True)
else:
self.room_class_relation = torch.nn.Parameter(torch.rand(num_rooms,num_rooms), requires_grad=True)
input_dim = feature_size + embedding_size
if not wo_instr_input:
input_dim = input_dim + hidden_size
if use_object:
input_dim = input_dim+300
self.input_mapping = nn.Linear(input_dim,hidden_size)
self.object_gcn = GCN_pre_define(300,process_num=num_gcn,max_degree=max_degree,short_cut=short_cut,load_adj_weight=load_object_relation_weight)
self.concate_room = concate_room
if use_room:
if concate_room:
self.action_selector = WhSoftDotAttention(hidden_size+hidden_size,visual_hidden_size+62)
else:
self.action_selector = WhSoftDotAttention(hidden_size+hidden_size,visual_hidden_size+repeat)
else:
self.action_selector = WhSoftDotAttention(hidden_size+hidden_size,visual_hidden_size)
self.visual_mlp = nn.Sequential(
nn.BatchNorm1d(feature_size),
nn.Linear(feature_size, visual_hidden_size),
nn.BatchNorm1d(visual_hidden_size),
nn.Dropout(dropout_ratio),
nn.ReLU())
self.visual_attention_layer = WhSoftDotAttention(hidden_size, visual_hidden_size)
def activate_classifier(self):
for p in self.view_room_classifier.parameters():
p.requires_grad=True
for p in self.text_room_classifier.parameters():
p.requires_grad=True
def forward(self,*args):
if self.search:
return self._forward_search(*args)
else:
return self._forward_train(*args)
def encode_label(self,labels):
embeddings = self.object_gcn.kg.embeddings
zero_tensor = try_cuda(torch.zeros(1,300))
embeddings = torch.cat([zero_tensor,embeddings],axis=0)
batch_size,label_num = labels.shape
if label_num < 5:
print('not enough label')
return try_cuda(torch.zeros(batch_size,5,300))
else:
labels = labels[:,:5]
origin_shape = labels.shape
labels = labels.reshape(-1)
embeding_features = embeddings[labels]
embeding_features = embeding_features.reshape((*origin_shape,-1))
return embeding_features
def get_label_sim(self,text_label_class,view_label_class,repeat=128):
if self.concate_room:
return torch.cat([text_label_class,view_label_class],axis=1)
if not self.room_relation_vec:
mid = torch.matmul(self.room_class_relation,text_label_class[:,:,None])
similarity = torch.bmm(view_label_class[:,None,:],mid)
similarity = similarity.squeeze()[:,None]
return similarity.expand(-1,128)
else:
mid =torch.matmul(text_label_class,self.room_class_relation)
return torch.bmm(view_label_class[:,None,:],mid.transpose(0,1)).squeeze()
def find_nearby(self,text_object_query):
if self.topk == 0:
return text_object_query
batch_size = text_object_query.shape[0]
embeddings = self.object_gcn.kg.embeddings
similarity = torch.mm(text_object_query,embeddings.transpose(0,1))
values,index = torch.topk(similarity,self.topk,dim=1)
values= torch.softmax(values,dim=1)
find_object = embeddings[index.reshape(-1)].reshape(batch_size,5,-1)
attn_object = find_object * values[:,:,None].expand(-1,-1,300)
attn_object = attn_object.sum(axis=1)
return attn_object,index
def is_search(self):
self.search = True
def not_search(self):
self.search = False
def _build_current_input(self,u_t_prev, all_u_t, visual_context,
h_0, ctx,object_label_set=None,view_label_set=None):
batch_size,action_count,_ = all_u_t.shape
ctx_pos = self.positional_encoding(ctx)#b,seq,512
text_obj_atten,alpha_atten_obj = self.object_text_attention_layer(h_0,ctx_pos,v=ctx)#(2)
text_room_atten,alpha_atten_room = self.room_text_attention_layer(h_0,ctx_pos,v=ctx)#(1)
text_atten = (text_obj_atten+text_room_atten)/2
text_alpha = (alpha_atten_obj+alpha_atten_room)/2
#xxx model
text_obj_query = self.object_map(text_obj_atten)
text_refined_object,object_index = self.find_nearby(text_obj_query)
text_room_class = self.text_room_classifier(text_room_atten)
text_hard_room_label = hard_softmax(text_room_class)
#yyy model
#common_sense part
object_label_features = self.object_gcn(object_label_set)
object_features,_ = self.object_attention_layer(text_refined_object,object_label_features)
encoded_room_feature = self.encode_label(view_label_set)
encoded_room_feature = encoded_room_feature.reshape(batch_size*action_count,-1)
view_room_class = self.view_room_classifier(encoded_room_feature)#room type classification
view_hard_room_label = hard_softmax(view_room_class)
text_hard_room_label_ = text_hard_room_label[:,None,:].expand(-1,action_count,-1).\
reshape(action_count*batch_size,-1)
if self.soft_room_label:
sim = self.get_label_sim(view_room_class,text_room_class[:,None,:].expand(-1,action_count,-1).\
reshape(action_count*batch_size,-1))
else:
sim = self.get_label_sim(text_hard_room_label_,view_hard_room_label)
sim = sim.reshape(batch_size,action_count,-1)
#ragular part
g_v = all_u_t.view(-1, self.feature_size)#b*ac, 2048+128z
g_v = self.visual_mlp(g_v).view(batch_size, action_count, -1)#b,ac,1024
attn_vision, _alpha_vision = self.visual_attention_layer(h_0, g_v, v=all_u_t)#b,2048+128 #b,seq,1
alpha_vision = self.sm(_alpha_vision)
concat_input = torch.cat([attn_vision,u_t_prev],axis=1)
if not self.wo_instr_input:
concat_input = torch.cat([text_atten,concat_input],axis=1)#b,2048+128+2048+128
if self.use_object:
concat_input = torch.cat([object_features,concat_input],axis=1)
_input_curent = self.input_mapping(concat_input)
return _input_curent,ctx_pos , text_atten, sim.reshape(batch_size,action_count,-1), \
g_v, attn_vision, text_alpha,alpha_vision,text_room_class,view_room_class, \
alpha_atten_obj, alpha_atten_room,object_index
def _forward_train(self,u_t_prev, all_u_t, visual_context, input_history, h_0, ctx,
ctx_mask,object_label_set=None,view_label_set=None):
import ipdb;ipdb.set_trace()
_input_curent,ctx_pos,text_atten,\
room_features,g_v,attn_vision,text_alpha,\
alpha_vision,text_room_class ,view_room_loss, alpha_atten_obj, alpha_atten_room, object_index = self._build_current_input(u_t_prev, all_u_t,
visual_context, h_0, ctx,
object_label_set,view_label_set)
input_curent = _input_curent[:,None,:]#b,512
ctx_mask = ctx_mask.bool()
if input_history is not None:
to_input = torch.cat([input_history,input_curent],axis=1)#b,seq+1,512
else:
to_input = input_curent
_to_input = self.positional_encoding(to_input)
ctx_input = ctx_pos.transpose(0,1)
_to_input = _to_input.transpose(0,1)
outputs = self.decoder(_to_input,ctx_input, memory_key_padding_mask=ctx_mask.bool())
outputs = outputs.transpose(0,1)
h_1 = outputs[:,-1,:]
c_1 = outputs[:,:-1,:]
c_1 = c_1.mean(axis=1)
# action_selector = torch.cat((attn_text, h_1),axis=1)#512,512
# _,alpha_action = self.action_attention_layer(action_selector,g_v)
if self.use_room:
g_v = torch.cat([g_v,room_features],axis=2)
alpha_action = self.action_prediction(text_atten,h_1,all_u_t,g_v)
return h_1,c_1,to_input,text_atten,attn_vision,text_alpha,alpha_action,alpha_vision,text_room_class, \
view_room_loss,self.sm(alpha_atten_obj),self.sm(alpha_atten_room),object_index
def _forward_search(self,u_t_prev, all_u_t, visual_context, input_history, history_length,h_0, ctx,
ctx_mask=None,object_label_set=None,view_label_set=None,gt_labels=None):
input_curent,ctx_pos,text_atten,room_features,\
g_v,attn_vision,text_alpha,alpha_vision,text_room_class ,view_room_loss,\
alpha_atten_obj,alpha_atten_room,object_index = self._build_current_input(u_t_prev, all_u_t,
visual_context, h_0, ctx,
object_label_set,view_label_set)
batch_size,action_count,_ = all_u_t.shape
batch_size = all_u_t.shape[0]
_,seq_len,_ = input_history.size()
to_input = input_history
to_input[np.arange(batch_size),history_length,:] = input_curent
tgt_mask = to_input != 0
tgt_mask = tgt_mask.sum(axis=2) != 0
c_mask = tgt_mask.clone()
tgt_mask = ~tgt_mask.bool()
_to_input = self.positional_encoding(to_input)
_to_input = _to_input.transpose(0,1)
ctx_input = ctx_pos.transpose(0,1)
ctx_mask = ctx_mask.bool()
outputs = self.decoder(_to_input,ctx_input,tgt_key_padding_mask=tgt_mask,memory_key_padding_mask=ctx_mask)
outputs = outputs.transpose(0,1)
h_1 = outputs[np.arange(batch_size),history_length,:]
c_mask[np.arange(batch_size),history_length] = 0
c_1 = outputs*c_mask[:,:,None].expand(batch_size,seq_len,512).float()
history_length = torch.tensor(history_length)
c_1 = c_1.sum(axis=1)/try_cuda(history_length[:,None].expand(batch_size,self.hidden_size)).clamp_min(1).float()
#action_selector = torch.cat((attn_text, h_1),axis=1)#512,512
#_,alpha_action = self.action_attention_layer(action_selector,g_v)
if self.use_room:
if gt_labels is not None:
text_labels = gt_labels[0]
view_labels = gt_labels[1]
text_labels_oh = try_cuda(torch.zeros(text_room_class.shape))
view_labels_oh = try_cuda(torch.zeros(view_room_loss.shape))
text_labels_oh.scatter_(1, text_labels[:,None], 1)
view_labels_oh.scatter(1,view_labels[:,None],1)
sim = self.get_label_sim(view_labels_oh,text_labels_oh[:,None,:].expand(-1,action_count,-1).\
reshape(action_count*batch_size,-1))
sim = sim.reshape(batch_size,action_count,-1)
room_features=sim
g_v = torch.cat([g_v,room_features],axis=2)
alpha_action = self.action_prediction(text_atten,h_1,all_u_t,g_v)
new_length = history_length+1
to_input = torch.cat([to_input,try_cuda(torch.zeros(batch_size ,1, self.hidden_size))],dim=1)
return h_1,c_1,to_input,new_length, text_atten,attn_vision,text_alpha,alpha_action,alpha_vision
def action_prediction(self,text_atten,h_1,all_u_t,g_v):
selector_hi = torch.cat([text_atten,h_1],axis=1)
_,alpha_action = self.action_selector(selector_hi,g_v)
return alpha_action
###############################################################################
# scorer models
###############################################################################