|
1
|
|
|
"""Introduces Lookup brick.""" |
|
2
|
|
|
from blocks.bricks import Initializable, Feedforward |
|
3
|
|
|
from blocks.bricks.base import application, lazy |
|
4
|
|
|
from blocks.roles import WEIGHT, add_role |
|
5
|
|
|
from blocks.utils import check_theano_variable, shared_floatx_nans |
|
6
|
|
|
|
|
7
|
|
|
|
|
8
|
|
|
class LookupTable(Initializable, Feedforward): |
|
9
|
|
|
"""Encapsulates representations of a range of integers. |
|
10
|
|
|
|
|
11
|
|
|
This brick can be used to embed integers, e.g. word indices, |
|
12
|
|
|
into a vector space. |
|
13
|
|
|
|
|
14
|
|
|
Parameters |
|
15
|
|
|
---------- |
|
16
|
|
|
length : int |
|
17
|
|
|
The size of the lookup table, or in other words, one plus the |
|
18
|
|
|
maximum index for which a representation is contained. |
|
19
|
|
|
dim : int |
|
20
|
|
|
The dimensionality of representations. |
|
21
|
|
|
|
|
22
|
|
|
Notes |
|
23
|
|
|
----- |
|
24
|
|
|
See :class:`.Initializable` for initialization parameters. |
|
25
|
|
|
|
|
26
|
|
|
""" |
|
27
|
|
|
has_bias = False |
|
28
|
|
|
|
|
29
|
|
|
@lazy(allocation=['length', 'dim']) |
|
30
|
|
|
def __init__(self, length, dim, **kwargs): |
|
31
|
|
|
super(LookupTable, self).__init__(**kwargs) |
|
32
|
|
|
self.length = length |
|
33
|
|
|
self.dim = dim |
|
34
|
|
|
|
|
35
|
|
|
@property |
|
36
|
|
|
def W(self): |
|
37
|
|
|
return self.parameters[0] |
|
38
|
|
|
|
|
39
|
|
|
def _allocate(self): |
|
40
|
|
|
self.parameters.append(shared_floatx_nans((self.length, self.dim), |
|
41
|
|
|
name='W')) |
|
42
|
|
|
add_role(self.parameters[-1], WEIGHT) |
|
43
|
|
|
|
|
44
|
|
|
@application(inputs=['indices'], outputs=['output']) |
|
45
|
|
|
def apply(self, indices): |
|
46
|
|
|
"""Perform lookup. |
|
47
|
|
|
|
|
48
|
|
|
Parameters |
|
49
|
|
|
---------- |
|
50
|
|
|
indices : :class:`~tensor.TensorVariable` |
|
51
|
|
|
The indices of interest. The dtype must be integer. |
|
52
|
|
|
|
|
53
|
|
|
Returns |
|
54
|
|
|
------- |
|
55
|
|
|
output : :class:`~tensor.TensorVariable` |
|
56
|
|
|
Representations for the indices of the query. Has :math:`k+1` |
|
57
|
|
|
dimensions, where :math:`k` is the number of dimensions of the |
|
58
|
|
|
`indices` parameter. The last dimension stands for the |
|
59
|
|
|
representation element. |
|
60
|
|
|
|
|
61
|
|
|
""" |
|
62
|
|
|
check_theano_variable(indices, None, ("int", "uint")) |
|
63
|
|
|
output_shape = [indices.shape[i] |
|
64
|
|
|
for i in range(indices.ndim)] + [self.dim] |
|
65
|
|
|
return self.W[indices.flatten()].reshape(output_shape) |
|
66
|
|
|
|
|
67
|
|
|
def get_dim(self, name): |
|
68
|
|
|
if name == 'output': |
|
69
|
|
|
return self.dim |
|
70
|
|
|
if name == 'indices': |
|
71
|
|
|
return 0 |
|
72
|
|
|
return super(LookupTable, self).get_dim(name) |
|
73
|
|
|
|
|
74
|
|
|
@property |
|
75
|
|
|
def input_dim(self): |
|
76
|
|
|
return 0 |
|
77
|
|
|
|
|
78
|
|
|
@input_dim.setter |
|
79
|
|
|
def input_dim(self, dim): |
|
80
|
|
|
if dim != 0: |
|
81
|
|
|
raise ValueError("LookupTable input must be integer") |
|
82
|
|
|
|
|
83
|
|
|
@property |
|
84
|
|
|
def output_dim(self): |
|
85
|
|
|
return self.dim |
|
86
|
|
|
|
|
87
|
|
|
@output_dim.setter |
|
88
|
|
|
def output_dim(self, dim): |
|
89
|
|
|
self.dim = dim |
|
90
|
|
|
|
mila-udem /
blocks
