-
Notifications
You must be signed in to change notification settings - Fork 0
/
model.py
955 lines (723 loc) · 35.8 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
import torch
import torch.nn as nn
import math
class InputEmbeddings(nn.Module):
"""
InputEmbeddings is a PyTorch module for creating input embeddings for a Transformer model.
This module combines a standard embedding layer with a scaling factor to produce
embeddings of a specified dimension (`d_model`). The scaling factor is the square
root of `d_model`, which helps in stabilizing the gradients during training.
Attributes:
d_model (int): The dimension of the embeddings.
vocab_size (int): The size of the vocabulary.
embedding (nn.Embedding): The embedding layer which converts input tokens to
dense vectors of size `d_model`.
Methods:
forward(x):
Applies the embedding layer to the input tensor `x` and scales the embeddings.
Parameters:
d_model (int): The dimension of the embeddings.
vocab_size (int): The size of the vocabulary.
"""
def __init__(self,
d_model: int,
vocab_size: int) -> None:
"""
Initializes the InputEmbeddings module.
Args:
d_model (int): The dimension of the embeddings.
vocab_size (int): The size of the vocabulary.
"""
super().__init__()
self.d_model = d_model
self.vocab_size = vocab_size
self.embedding = nn.Embedding(vocab_size, d_model)
def forward(self,
x):
"""
Applies the embedding layer to the input tensor `x` and scales the embeddings.
Args:
x (Tensor): Input tensor of token indices with shape (batch_size, sequence_length).
Returns:
Tensor: Scaled embeddings with shape (batch_size, sequence_length, d_model).
"""
return self.embedding(x) * math.sqrt(self.d_model)
class PositionalEncoding(nn.Module):
"""
PositionalEncoding is a PyTorch module that adds positional information to the input embeddings.
This module creates a matrix of positional encodings which are added to the input embeddings
to provide the model with information about the position of each token in the sequence.
It also applies dropout to the resulting tensor.
Attributes:
d_model (int): The dimension of the embeddings.
seq_len (int): The maximum length of the input sequences.
dropout (nn.Dropout): Dropout layer for regularization.
pe (Tensor): Positional encoding matrix of shape (1, seq_len, d_model).
Methods:
forward(x):
Adds positional encodings to the input tensor `x` and applies dropout.
Parameters:
d_model (int): The dimension of the embeddings.
seq_len (int): The maximum length of the input sequences.
dropout (float): The dropout rate to apply to the output.
"""
def __init__(self,
d_model: int,
seq_len: int,
dropout: float) -> None:
"""
Initializes the PositionalEncoding module.
Args:
d_model (int): The dimension of the embeddings.
seq_len (int): The maximum length of the input sequences.
dropout (float): The dropout rate to apply to the output.
"""
super().__init__()
self.d_model = d_model
self.seq_len = seq_len
self.dropout = nn.Dropout(dropout)
# Matrix Of Shape (seq_len, d_model)
pe = torch.zeros(seq_len, d_model)
# Vector Of Shape (seq_len, 1)
position = torch.arange(0, seq_len, dtype=torch.float).unsqueeze(1)
div_term = torch.exp(torch.arange(0, d_model, 2).float() * (-math.log(10000.0) / d_model))
# Sine Function applied on Even Positions
pe[:, 0::2] = torch.sin(position * div_term)
# Cosine Function applied on Odd Positions
pe[:, 1::2] = torch.cos(position * div_term)
# Expand Dimension to include Batches (1, seq_len, d_model)
pe = pe.unsqueeze(0)
# Register the Tensor in Buffer to Save Along the Model
self.register_buffer('pe', pe)
def forward(self,
x):
"""
Adds positional encodings to the input tensor `x` and applies dropout.
Args:
x (Tensor): Input tensor of shape (batch_size, seq_len, d_model).
Returns:
Tensor: The input tensor with added positional encodings and applied dropout,
of shape (batch_size, seq_len, d_model).
"""
x = x + (self.pe[:, :x.shape[1], :]).requires_grad_(False)
return self.dropout(x)
class LayerNormalization(nn.Module):
"""
LayerNormalization is a PyTorch module that applies layer normalization to its inputs.
This module normalizes the inputs across the features for each data point, ensuring that
the mean and variance of the inputs remain constant, which helps in stabilizing and speeding
up the training process. It also includes learnable scaling (`alpha`) and shifting (`bias`)
parameters to allow the model to undo normalization if necessary.
Attributes:
eps (float): A small value added to the denominator for numerical stability.
alpha (nn.Parameter): Learnable scaling parameter, initialized to ones.
bias (nn.Parameter): Learnable shifting parameter, initialized to zeros.
Methods:
forward(x):
Applies layer normalization to the input tensor `x`.
Parameters:
eps (float): A small value added to the denominator for numerical stability. Default is 1e-6.
"""
def __init__(self,
features: int,
eps: float = 10**-6) -> None:
"""
Initializes the LayerNormalization module.
Args:
eps (float): A small value added to the denominator for numerical stability. Default is 1e-6.
"""
super().__init__()
self.eps = eps
self.alpha = nn.Parameter(torch.ones(features)) # Multiplied
self.bias = nn.Parameter(torch.zeros(features)) # Added
def forward(self,
x):
"""
Applies layer normalization to the input tensor `x`.
Args:
x (Tensor): Input tensor of shape (batch_size, ..., feature_size).
Returns:
Tensor: The normalized tensor with the same shape as `x`.
"""
mean = x.mean(dim = -1, keepdim=True)
std = x.std(dim = -1, keepdim=True)
return self.alpha * (x - mean)/ (std + self.eps) + self.bias
class FeedForwardBlock(nn.Module):
"""
FeedForwardBlock is a PyTorch module that implements a feedforward neural network block
used in Transformer models.
This module consists of two linear transformations with a ReLU activation in between,
followed by dropout for regularization. It is typically used after the self-attention
mechanism in each layer of the Transformer to further process the data.
Attributes:
linear_1 (nn.Linear): The first linear transformation layer (d_model to d_ff).
dropout (nn.Dropout): Dropout layer for regularization.
linear_2 (nn.Linear): The second linear transformation layer (d_ff to d_model).
Methods:
forward(x):
Applies the feedforward block to the input tensor `x`.
Parameters:
d_model (int): The dimension of the input and output features.
d_ff (int): The dimension of the inner layer.
dropout (float): The dropout rate to apply after the first linear transformation.
"""
def __init__(self,
d_model: int,
d_ff: int,
dropout: float)-> None:
"""
Initializes the FeedForwardBlock module.
Args:
d_model (int): The dimension of the input and output features.
d_ff (int): The dimension of the inner layer.
dropout (float): The dropout rate to apply after the first linear transformation.
"""
super().__init__()
# Weights are already true in the Linear Layer
self.linear_1 = nn.Linear(d_model, d_ff) # W1 + b1
self.dropout = nn.Dropout(dropout)
self.linear_2 = nn.Linear(d_ff, d_model) # W2 + b2
def forward(self,
x):
"""
Applies the feedforward block to the input tensor `x`.
Args:
x (Tensor): Input tensor of shape (batch_size, seq_len, d_model).
Returns:
Tensor: Output tensor of shape (batch_size, seq_len, d_model) after applying
two linear transformations with ReLU activation and dropout in between.
"""
return self.linear_2(self.dropout(torch.relu(self.linear_1(x))))
class MultiHeadAttention(nn.Module):
"""
MultiHeadAttention is a PyTorch module that implements the multi-head attention mechanism
used in Transformer models.
This module applies multiple attention heads to the input sequences, allowing the model to
focus on different parts of the input for each head. It consists of linear transformations
for queries, keys, and values, as well as a final linear transformation for the output.
Attributes:
d_model (int): The dimension of the input and output features.
h (int): The number of attention heads.
d_k (int): The dimension of the queries and keys for each head (d_model // h).
w_q (nn.Linear): Linear layer for transforming inputs to queries.
w_k (nn.Linear): Linear layer for transforming inputs to keys.
w_v (nn.Linear): Linear layer for transforming inputs to values.
w_o (nn.Linear): Linear layer for transforming concatenated outputs of all heads.
dropout (nn.Dropout): Dropout layer for regularization.
Methods:
attention(query, key, value, mask, dropout):
Computes scaled dot-product attention.
forward(q, k, v, mask):
Applies the multi-head attention mechanism to the input tensors.
Parameters:
d_model (int): The dimension of the input and output features.
h (int): The number of attention heads.
dropout (float): The dropout rate to apply to the attention weights.
"""
def __init__(self,
d_model: int,
h: int,
dropout: float) -> None:
"""
Initializes the MultiHeadAttention module.
Args:
d_model (int): The dimension of the input and output features.
h (int): The number of attention heads.
dropout (float): The dropout rate to apply to the attention weights.
"""
super().__init__()
self.d_model = d_model
self.h = h
assert d_model % h == 0, "d_model not divisible by h"
self.d_k = d_model // h
self.w_q = nn.Linear(d_model, d_model, bias=False) # Wq
self.w_k = nn.Linear(d_model, d_model, bias=False) # Wk
self.w_v = nn.Linear(d_model, d_model, bias=False) # Wv
self.w_o = nn.Linear(d_model, d_model, bias=False) # Wo
self.dropout = nn.Dropout(dropout)
@staticmethod
def attention(query,
key,
value,
mask,
dropout: nn.Dropout):
"""
Computes scaled dot-product attention.
Args:
query (Tensor): Query tensor of shape (batch_size, h, seq_len, d_k).
key (Tensor): Key tensor of shape (batch_size, h, seq_len, d_k).
value (Tensor): Value tensor of shape (batch_size, h, seq_len, d_k).
mask (Tensor): Mask tensor to prevent attention to certain positions.
dropout (nn.Dropout): Dropout layer to apply to the attention scores.
Returns:
Tensor: Output tensor after applying attention mechanism.
Tensor: Attention scores.
"""
d_k = query.shape[-1]
# (Batch, h, seq_len, d_k) --> (Batch, h , seq_len, seq_len)
attention_scores = (query @ key.transpose(-2, -1))/math.sqrt(d_k)
if mask is not None:
attention_scores.masked_fill_(mask == 0, -1e9)
attention_scores = attention_scores.softmax(dim = -1) # (Batch, h seq_len, seq_len)
if dropout is not None:
attention_scores = dropout(attention_scores)
return (attention_scores @ value), attention_scores
def forward(self,
q,
k,
v,
mask):
"""
Applies the multi-head attention mechanism to the input tensors.
Args:
q (Tensor): Query tensor of shape (batch_size, seq_len, d_model).
k (Tensor): Key tensor of shape (batch_size, seq_len, d_model).
v (Tensor): Value tensor of shape (batch_size, seq_len, d_model).
mask (Tensor): Mask tensor to prevent attention to certain positions.
Returns:
Tensor: Output tensor of shape (batch_size, seq_len, d_model).
"""
query = self.w_q(q) # (batch, seq_len, d_model) --> (batch, seq_len, d_model)
key = self.w_k(k) # (batch, seq_len, d_model) --> (batch, seq_len, d_model)
value = self.w_v(v) # (batch, seq_len, d_model) --> (batch, seq_len, d_model)
# (batch, seq_len, d_model) --> (batch, seq_len, h, d_k) --> (batch, h, seq_len, d_k)
query = query.view(query.shape[0], query.shape[1], self.h, self.d_k).transpose(1, 2)
key = key.view(key.shape[0], key.shape[1], self.h, self.d_k).transpose(1, 2)
value = value.view(value.shape[0], value.shape[1], self.h, self.d_k).transpose(1, 2)
x, self.attention_scores = MultiHeadAttention.attention(query, key, value, mask, self.dropout)
# (Batch, h, seq_len, d_k) --> (Batch, seq_len, h, d_k) --> (Batch, seq_len, d_model)
x = x.transpose(1, 2).contiguous().view(x.shape[0], -1, self.h * self.d_k)
# (Batch, seq_len, d_model) --> (Batch, seq_len, d_model)
return self.w_o(x)
class ResidualConnection(nn.Module):
"""
ResidualConnection is a PyTorch module that implements a residual connection with
layer normalization and dropout.
This module is used in Transformer models to add the input of a sublayer to its output,
which helps in training deep networks by mitigating the vanishing gradient problem.
It includes layer normalization before the sublayer and dropout after the sublayer.
Attributes:
dropout (nn.Dropout): Dropout layer for regularization.
norm (LayerNormalization): Layer normalization applied before the sublayer.
Methods:
forward(x, sublayer):
Applies the residual connection to the input tensor `x` and the output of the `sublayer`.
Parameters:
dropout (float): The dropout rate to apply after the sublayer.
"""
def __init__(self,
features: int,
dropout: float) -> None:
"""
Initializes the ResidualConnection module.
Args:
dropout (float): The dropout rate to apply after the sublayer.
"""
super().__init__()
self.dropout = nn.Dropout(dropout)
self.norm = LayerNormalization(features)
def forward(self,
x,
sublayer):
"""
Applies the residual connection to the input tensor `x` and the output of the `sublayer`.
Args:
x (Tensor): Input tensor of shape (batch_size, seq_len, d_model).
sublayer (Callable): A sublayer function or module that takes `x` as input and returns a tensor of the same shape.
Returns:
Tensor: Output tensor of shape (batch_size, seq_len, d_model) after applying layer normalization,
the sublayer, dropout, and adding the input `x`.
"""
return x + self.dropout(sublayer(self.norm(x)))
class EncoderBlock(nn.Module):
"""
EncoderBlock is a PyTorch module that implements a single block of the encoder in a Transformer model.
This module consists of a multi-head self-attention mechanism followed by a position-wise
feed-forward neural network, each with residual connections and layer normalization.
Attributes:
self_attention_block (MultiHeadAttention): Multi-head self-attention mechanism.
feed_forward_block (FeedForwardBlock): Position-wise feed-forward neural network.
residual_connections (nn.ModuleList): List of residual connections for the self-attention and feed-forward blocks.
Methods:
forward(x, src_mask):
Applies the encoder block to the input tensor `x` with the provided source mask `src_mask`.
Parameters:
self_attention_block (MultiHeadAttention): Multi-head self-attention mechanism.
feed_forward_block (FeedForwardBlock): Position-wise feed-forward neural network.
dropout (float): The dropout rate to apply in the residual connections.
"""
def __init__(self,
features: int,
self_attention_block: MultiHeadAttention,
feed_forward_block: FeedForwardBlock,
dropout: float) -> None:
"""
Initializes the EncoderBlock module.
Args:
self_attention_block (MultiHeadAttention): Multi-head self-attention mechanism.
feed_forward_block (FeedForwardBlock): Position-wise feed-forward neural network.
dropout (float): The dropout rate to apply in the residual connections.
"""
super().__init__()
self.self_attention_block = self_attention_block
self.feed_forward_block = feed_forward_block
self.residual_connections = nn.ModuleList([ResidualConnection(features, dropout) for _ in range(2)])
def forward(self,
x,
src_mask):
"""
Applies the encoder block to the input tensor `x` with the provided source mask `src_mask`.
Args:
x (Tensor): Input tensor of shape (batch_size, seq_len, d_model).
src_mask (Tensor): Source mask tensor to prevent attention to certain positions.
Returns:
Tensor: Output tensor of shape (batch_size, seq_len, d_model) after applying the self-attention
and feed-forward blocks with residual connections.
"""
x = self.residual_connections[0](x, lambda x: self.self_attention_block(x, x, x, src_mask))
x = self.residual_connections[1](x, self.feed_forward_block)
return x
class Encoder(nn.Module):
"""
Encoder is a PyTorch module that implements the encoder component of a Transformer model.
The Encoder consists of a stack of encoder blocks, each containing a multi-head self-attention mechanism
followed by a position-wise feed-forward neural network, with residual connections and layer normalization.
Attributes:
layers (nn.ModuleList): List of encoder blocks.
norm (LayerNormalization): Layer normalization applied to the output of the final encoder block.
Methods:
forward(x, mask):
Applies the stack of encoder blocks to the input tensor `x` with the provided mask `mask`.
Parameters:
layers (nn.ModuleList): List of encoder blocks.
"""
def __init__(self,
features: int,
layers: nn.ModuleList) -> None:
"""
Initializes the Encoder module.
Args:
layers (nn.ModuleList): List of encoder blocks.
"""
super().__init__()
self.layers = layers
self.norm = LayerNormalization(features)
def forward(self,
x,
mask):
"""
Applies the stack of encoder blocks to the input tensor `x` with the provided mask `mask`.
Args:
x (Tensor): Input tensor of shape (batch_size, seq_len, d_model).
mask (Tensor): Mask tensor to prevent attention to certain positions.
Returns:
Tensor: Output tensor of shape (batch_size, seq_len, d_model) after applying the encoder blocks
and layer normalization.
"""
for layer in self.layers:
x = layer(x, mask)
return self.norm(x)
class DecoderBlock(nn.Module):
"""
DecoderBlock is a PyTorch module that implements a single block of the decoder in a Transformer model.
This module consists of a self-attention mechanism, a cross-attention mechanism that attends to
the encoder's output, and a position-wise feed-forward neural network, each with residual connections
and layer normalization.
Attributes:
self_attention_block (MultiHeadAttention): Multi-head self-attention mechanism.
cross_attention_block (MultiHeadAttention): Multi-head cross-attention mechanism for attending to encoder outputs.
feed_forward_block (FeedForwardBlock): Position-wise feed-forward neural network.
residual_connections (nn.ModuleList): List of residual connections for the self-attention, cross-attention,
and feed-forward blocks.
Methods:
forward(x, encoder_output, src_mask, tgt_mask):
Applies the decoder block to the input tensor `x` with the provided encoder output and masks.
Parameters:
self_attention_block (MultiHeadAttention): Multi-head self-attention mechanism.
cross_attention_block (MultiHeadAttention): Multi-head cross-attention mechanism for attending to encoder outputs.
feed_forward_block (FeedForwardBlock): Position-wise feed-forward neural network.
dropout (float): The dropout rate to apply in the residual connections.
"""
def __init__(self,
features: int,
self_attention_block: MultiHeadAttention,
cross_attention_block: MultiHeadAttention,
feed_forward_block: FeedForwardBlock,
dropout: float) -> None:
"""
Initializes the DecoderBlock module.
Args:
self_attention_block (MultiHeadAttention): Multi-head self-attention mechanism.
cross_attention_block (MultiHeadAttention): Multi-head cross-attention mechanism for attending to encoder outputs.
feed_forward_block (FeedForwardBlock): Position-wise feed-forward neural network.
dropout (float): The dropout rate to apply in the residual connections.
"""
super().__init__()
self.self_attention_block = self_attention_block
self.cross_attention_block = cross_attention_block
self.feed_forward_block = feed_forward_block
self.residual_connections = nn.ModuleList([ResidualConnection(features, dropout) for _ in range(3)])
def forward(self,
x,
encoder_output,
src_mask,
tgt_mask):
"""
Applies the decoder block to the input tensor `x` with the provided encoder output and masks.
Args:
x (Tensor): Input tensor of shape (batch_size, tgt_seq_len, d_model).
encoder_output (Tensor): Output tensor from the encoder of shape (batch_size, src_seq_len, d_model).
src_mask (Tensor): Source mask tensor to prevent attention to certain positions in the encoder output.
tgt_mask (Tensor): Target mask tensor to prevent attention to certain positions in the target input.
Returns:
Tensor: Output tensor of shape (batch_size, tgt_seq_len, d_model) after applying the self-attention,
cross-attention, and feed-forward blocks with residual connections.
"""
x = self.residual_connections[0](x, lambda x: self.self_attention_block(x, x, x, tgt_mask))
x = self.residual_connections[1](x, lambda x: self.cross_attention_block(x, encoder_output, encoder_output, src_mask))
x = self.residual_connections[2](x, self.feed_forward_block)
return x
class Decoder(nn.Module):
"""
Decoder is a PyTorch module that implements the decoder component of a Transformer model.
The Decoder consists of a stack of decoder blocks, each containing a self-attention mechanism,
a cross-attention mechanism that attends to the encoder's output, and a position-wise feed-forward
neural network, with residual connections and layer normalization.
Attributes:
layers (nn.ModuleList): List of decoder blocks.
norm (LayerNormalization): Layer normalization applied to the output of the final decoder block.
Methods:
forward(x, encoder_output, src_mask, tgt_mask):
Applies the stack of decoder blocks to the input tensor `x` with the provided encoder output and masks.
Parameters:
layers (nn.ModuleList): List of decoder blocks.
"""
def __init__(self,
features: int,
layers: nn.ModuleList) -> None:
"""
Initializes the Decoder module.
Args:
layers (nn.ModuleList): List of decoder blocks.
"""
super().__init__()
self.layers = layers
self.norm = LayerNormalization(features)
def forward(self,
x,
encoder_output,
src_mask,
tgt_mask):
"""
Applies the stack of decoder blocks to the input tensor `x` with the provided encoder output and masks.
Args:
x (Tensor): Input tensor of shape (batch_size, tgt_seq_len, d_model).
encoder_output (Tensor): Output tensor from the encoder of shape (batch_size, src_seq_len, d_model).
src_mask (Tensor): Source mask tensor to prevent attention to certain positions in the encoder output.
tgt_mask (Tensor): Target mask tensor to prevent attention to certain positions in the target input.
Returns:
Tensor: Output tensor of shape (batch_size, tgt_seq_len, d_model) after applying the decoder blocks
and layer normalization.
"""
for layer in self.layers:
x = layer(x, encoder_output, src_mask, tgt_mask)
return self.norm(x)
class ProjectionLayer(nn.Module):
"""
ProjectionLayer is a PyTorch module that implements the projection layer of a Transformer model.
The ProjectionLayer linearly transforms the input tensor from the dimension of `d_model` to the
dimension of `vocab_size`, followed by a log softmax operation along the last dimension.
Attributes:
proj (nn.Linear): Linear transformation layer.
Methods:
forward(x):
Applies the projection layer to the input tensor `x`.
Parameters:
d_model (int): Dimensionality of the input tensor.
vocab_size (int): Size of the vocabulary, i.e., the number of output classes.
"""
def __init__(self,
d_model: int,
vocab_size: int) -> None:
"""
Initializes the ProjectionLayer module.
Args:
d_model (int): Dimensionality of the input tensor.
vocab_size (int): Size of the vocabulary, i.e., the number of output classes.
"""
super().__init__()
self.proj = nn.Linear(d_model,
vocab_size)
def forward(self,
x):
"""
Applies the projection layer to the input tensor `x`.
Args:
x (Tensor): Input tensor of shape (batch_size, seq_len, d_model).
Returns:
Tensor: Output tensor of shape (batch_size, seq_len, vocab_size) after applying the linear transformation
followed by log softmax operation along the last dimension.
"""
# (Batch, seq_len, d_model) --> (Batch, seq_len, vocab_size)
return self.proj(x)
class Transformer(nn.Module):
"""
Transformer is a PyTorch module that implements the Transformer model.
The Transformer model consists of an encoder, a decoder, input embeddings, positional encodings,
and a projection layer. It is designed for sequence-to-sequence tasks such as machine translation.
Attributes:
encoder (Encoder): Encoder component of the Transformer.
decoder (Decoder): Decoder component of the Transformer.
src_embed (InputEmbeddings): Input embeddings for the source sequence.
tgt_embed (InputEmbeddings): Input embeddings for the target sequence.
src_pos (PositionalEncoding): Positional encodings for the source sequence.
tgt_pos (PositionalEncoding): Positional encodings for the target sequence.
projection_layer (ProjectionLayer): Projection layer to generate output probabilities.
Methods:
encode(src, src_mask):
Encodes the source sequence using the encoder.
decode(src_mask, encoder_output, tgt, tgt_mask):
Decodes the target sequence using the decoder and encoder output.
project(x):
Projects the output tensor `x` to obtain the final output probabilities.
Parameters:
encoder (Encoder): Encoder component of the Transformer.
decoder (Decoder): Decoder component of the Transformer.
src_embed (InputEmbeddings): Input embeddings for the source sequence.
tgt_embed (InputEmbeddings): Input embeddings for the target sequence.
src_pos (PositionalEncoding): Positional encodings for the source sequence.
tgt_pos (PositionalEncoding): Positional encodings for the target sequence.
projection_layer (ProjectionLayer): Projection layer to generate output probabilities.
"""
def __init__(self,
encoder: Encoder,
decoder: Decoder,
src_embed: InputEmbeddings,
tgt_embed: InputEmbeddings,
src_pos: PositionalEncoding,
tgt_pos: PositionalEncoding,
projection_layer: ProjectionLayer) -> None:
"""
Initializes the Transformer module.
Args:
encoder (Encoder): Encoder component of the Transformer.
decoder (Decoder): Decoder component of the Transformer.
src_embed (InputEmbeddings): Input embeddings for the source sequence.
tgt_embed (InputEmbeddings): Input embeddings for the target sequence.
src_pos (PositionalEncoding): Positional encodings for the source sequence.
tgt_pos (PositionalEncoding): Positional encodings for the target sequence.
projection_layer (ProjectionLayer): Projection layer to generate output probabilities.
"""
super().__init__()
self.encoder = encoder
self.decoder = decoder
self.src_embed = src_embed
self.tgt_embed = tgt_embed
self.src_pos = src_pos
self.tgt_pos = tgt_pos
self.projection_layer = projection_layer
def encode(self,
src,
src_mask):
"""
Encodes the source sequence using the encoder.
Args:
src (Tensor): Source sequence tensor.
src_mask (Tensor): Mask tensor for the source sequence.
Returns:
Tensor: Encoded representation of the source sequence.
"""
src = self.src_embed(src)
src = self.src_pos(src)
return self.encoder(src,
src_mask)
def decode(self,
encoder_output: torch.Tensor,
src_mask: torch.Tensor,
tgt: torch.Tensor,
tgt_mask: torch.Tensor):
"""
Decodes the target sequence using the decoder and encoder output.
Args:
src_mask (Tensor): Mask tensor for the source sequence.
encoder_output (Tensor): Output tensor from the encoder.
tgt (Tensor): Target sequence tensor.
tgt_mask (Tensor): Mask tensor for the target sequence.
Returns:
Tensor: Decoded representation of the target sequence.
"""
tgt = self.tgt_embed(tgt)
tgt = self.tgt_pos(tgt)
return self.decoder(tgt,
encoder_output,
src_mask,
tgt_mask)
def project(self,
x):
"""
Projects the output tensor `x` to obtain the final output probabilities.
Args:
x (Tensor): Input tensor to be projected.
Returns:
Tensor: Output tensor after projection.
"""
return self.projection_layer(x)
def build_transformer(src_vocab_size: int,
tgt_vocab_size: int,
src_seq_len: int,
tgt_seq_len: int,
d_model: int = 512,
N: int = 6,
h: int = 8,
dropout: float = 0.1,
d_ff: int = 2048) -> Transformer:
"""
Builds and initializes a Transformer model for sequence-to-sequence tasks.
The function constructs and initializes the components of a Transformer model
based on the provided parameters.
Args:
src_vocab_size (int): Size of the source vocabulary.
tgt_vocab_size (int): Size of the target vocabulary.
src_seq_len (int): Length of the source sequence.
tgt_seq_len (int): Length of the target sequence.
d_model (int, optional): Dimensionality of the model. Defaults to 512.
N (int, optional): Number of encoder and decoder blocks. Defaults to 6.
h (int, optional): Number of attention heads. Defaults to 8.
dropout (float, optional): Dropout probability. Defaults to 0.1.
d_ff (int, optional): Dimensionality of the feed-forward network. Defaults to 2048.
Returns:
Transformer: Initialized Transformer model.
"""
# Create The Embedding Layers
src_embed = InputEmbeddings(d_model, src_vocab_size)
tgt_embed = InputEmbeddings(d_model, tgt_vocab_size)
# Create The Positional Encoding Layers
src_pos = PositionalEncoding(d_model, src_seq_len, dropout)
tgt_pos = PositionalEncoding(d_model, tgt_seq_len, dropout)
# Create The Encoder Blocks
encoder_blocks = []
for _ in range(N):
encoder_self_attention_block = MultiHeadAttention(d_model, h, dropout)
encoder_feed_forward_block = FeedForwardBlock(d_model, d_ff, dropout)
encoder_block = EncoderBlock(d_model, encoder_self_attention_block, encoder_feed_forward_block, dropout)
encoder_blocks.append(encoder_block)
# Create The Decoder Blocks
decoder_blocks = []
for _ in range(N):
decoder_self_attention_block = MultiHeadAttention(d_model, h, dropout)
decoder_cross_attention_block = MultiHeadAttention(d_model, h, dropout)
decoder_feed_forward_block = FeedForwardBlock(d_model, d_ff, dropout)
decoder_block = DecoderBlock(d_model, decoder_self_attention_block, decoder_cross_attention_block, decoder_feed_forward_block, dropout)
decoder_blocks.append(decoder_block)
# Create The Encoder And Decoder
encoder = Encoder(d_model, nn.ModuleList(encoder_blocks))
decoder = Decoder(d_model, nn.ModuleList(decoder_blocks))
# Create The Projection Layer
projection_layer = ProjectionLayer(d_model, tgt_vocab_size)
# Create The Transformer
transformer = Transformer(encoder, decoder, src_embed, tgt_embed, src_pos, tgt_pos, projection_layer)
# Initialize The Parameters
for p in transformer.parameters():
if p.dim() > 1:
nn.init.xavier_uniform_(p)
return transformer