Mercurial > pylearn
annotate nnet_ops.py @ 67:810a8e3c85e1
fixed horrible memory leak from crossentropy...
author | bergstra@is23.m |
---|---|
date | Fri, 18 Apr 2008 03:35:58 -0400 |
parents | 1b152f46ad0c |
children | 315eb36ff954 |
rev | line source |
---|---|
24 | 1 import theano |
2 from theano import tensor, gof, scalar | |
3 import numpy | |
4 | |
5 class ScalarSigmoid(scalar.UnaryScalarOp): | |
6 def impl(self, x): | |
7 return 1.0 / (1 + numpy.exp(-x)) | |
8 def grad(self, (x,), (gz,)): | |
9 return gz * scalar_sigmoid(x) * (1.0 - scalar_sigmoid(x)), | |
10 def c_foreach(self, (x,), (z,)): | |
11 return "%(z)s = 1.0 / (1 + exp(-%(x)s));" % locals() | |
12 scalar_sigmoid = gof.op.constructor(ScalarSigmoid) | |
13 Sigmoid, sigmoid, SigmoidInplace, sigmoid_inplace \ | |
14 = theano.tensor.broadcast(ScalarSigmoid, 'Sigmoid') | |
15 | |
16 | |
17 | |
18 class CrossentropySoftmax1Hot(gof.op.Op): | |
19 """A special compound Op for the output of neural-net classifiers. | |
20 | |
21 This Op has two outputs: | |
22 - KL(softmax(x), y) | |
23 - softmax(x) | |
24 | |
25 x[i] is assumed to be a dense vector | |
26 softmax(x[i]) is the i'th distribution over len(x[i]) options | |
27 y[i] is an integer index, encoding a 1-hot distribution | |
28 | |
29 """ | |
30 nin=2 | |
31 nout=2 | |
30
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
32 def __init__(self, x, b, y_idx, **kwargs): |
24 | 33 x = tensor._as_tensor(x) |
30
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
34 b = tensor._as_tensor(b) |
24 | 35 y_idx = tensor._as_tensor(y_idx) |
30
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
36 if len(x.broadcastable) != 2 \ |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
37 or x.dtype not in ['float32', 'float64']: |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
38 raise ValueError('x must be 2-d tensor of floats') |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
39 if len(b.broadcastable) != 1 \ |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
40 or x.dtype not in ['float32', 'float64']: |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
41 raise ValueError('x must be 1-d tensor of floats') |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
42 if len(y_idx.broadcastable) != 1 \ |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
43 or y_idx.dtype not in ['int32', 'int64']: |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
44 raise ValueError('x must be 1-d tensor of ints') |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
45 |
24 | 46 # TODO: Is this correct? It used to be y, not y_idx |
47 nll = tensor.Tensor(x.dtype, y_idx.broadcastable) | |
48 # nll = Tensor(x.dtype, y.broadcastable) | |
49 sm = tensor.Tensor(x.dtype, x.broadcastable) | |
30
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
50 self.inputs = [x, b, y_idx] |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
51 self.outputs = [nll, sm] |
24 | 52 def perform(self): |
30
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
53 x, b, y_idx = [i.data for i in self.inputs] |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
54 if b.shape[0] != x.shape[1]: |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
55 raise ValueError('b must have same shape as x[0]') |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
56 |
24 | 57 sm = numpy.zeros_like(x) # softmax |
58 nll = numpy.zeros(x.shape[0]) #nll(y | softmax(x)) | |
59 for i in xrange(sm.shape[0]): | |
30
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
60 row = x[i] + b |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
61 sm[i] = numpy.exp(row - numpy.max(row)) #softmax |
24 | 62 sm[i] *= 1.0 / numpy.sum(sm[i]) #vector scale |
63 nll[i] = -numpy.log( sm[i, y_idx[i]]) #cross-entropy | |
64 self.outputs[0].data = nll | |
65 self.outputs[1].data = sm | |
30
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
66 def grad(self, (x, b, y_idx), (g_nll, g_sm)): |
24 | 67 if g_sm is not None: |
68 raise NotImplementedError() | |
30
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
69 nll, sm = crossentropy_softmax_1hot(x, b, y_idx) |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
70 dx = CrossentropySoftmax1HotDx(g_nll, sm, y_idx).outputs[0] |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
71 db = tensor.Sum(dx, axis = [0]).outputs[0] |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
72 return dx, db, None |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
73 |
67
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
74 def c_headers(self): return ['<iostream>'] |
30
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
75 def c_code(self, (x, b, y_idx), (nll, sm), sub): |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
76 # this implementation was lifted from |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
77 # /u/bergstrj/cvs/bergstrj/src/feb07/nn.cxx |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
78 |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
79 #TODO: put this into a templated function, in the support code |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
80 #TODO: declare the max of each row as an Op output |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
81 |
32 | 82 #TODO: set error messages for failures in this code |
83 | |
30
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
84 return """ |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
85 npy_intp* Nx = %(x)s->dimensions; |
34 | 86 |
67
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
87 if (%(x)s->nd != 2) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
88 { |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
89 PyErr_SetString(PyExc_ValueError, "a not 2d tensor"); |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
90 %(fail)s; |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
91 } |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
92 if (%(b)s->nd != 1) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
93 { |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
94 PyErr_SetString(PyExc_ValueError, "b not 1d tensor"); |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
95 %(fail)s; |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
96 } |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
97 if (%(y_idx)s->nd != 1) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
98 { |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
99 PyErr_SetString(PyExc_ValueError, "y_idx not 1d tensor"); |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
100 %(fail)s; |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
101 } |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
102 if (%(x)s->descr->type_num != PyArray_DOUBLE) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
103 { |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
104 PyErr_SetString(PyExc_TypeError, "a not float64"); |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
105 %(fail)s; |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
106 } |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
107 if (%(b)s->descr->type_num != PyArray_DOUBLE) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
108 { |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
109 PyErr_SetString(PyExc_TypeError, "b not float64"); |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
110 %(fail)s; |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
111 } |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
112 if (%(y_idx)s->descr->type_num != PyArray_INT64) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
113 { |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
114 PyErr_SetString(PyExc_TypeError, "y_idx not int64"); |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
115 %(fail)s; |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
116 } |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
117 if ((%(x)s->dimensions[1] != %(b)s->dimensions[0]) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
118 || (%(x)s->dimensions[0] != %(y_idx)s->dimensions[0])) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
119 { |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
120 PyErr_SetString(PyExc_ValueError, "dimension mismatch in arguments"); |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
121 %(fail)s; |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
122 } |
34 | 123 |
67
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
124 if ((NULL == %(nll)s) //initial condition |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
125 || (%(nll)s->dimensions[0] != %(y_idx)s->dimensions[0])) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
126 { |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
127 if (NULL != %(nll)s) Py_XDECREF(%(nll)s); |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
128 %(nll)s = (PyArrayObject*)PyArray_SimpleNew(1, PyArray_DIMS(%(y_idx)s), type_num_%(x)s); |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
129 if(!%(nll)s) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
130 { |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
131 PyErr_SetString(PyExc_MemoryError, "failed to alloc nll output"); |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
132 %(fail)s; |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
133 } |
34 | 134 } |
67
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
135 if ((NULL == %(sm)s) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
136 || (%(sm)s->dimensions[0] != %(x)s->dimensions[0]) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
137 || (%(sm)s->dimensions[1] != %(x)s->dimensions[1])) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
138 { |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
139 if (NULL != %(sm)s) Py_XDECREF(%(sm)s); |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
140 %(sm)s = (PyArrayObject*)PyArray_SimpleNew(2, PyArray_DIMS(%(x)s), type_num_%(x)s); |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
141 if(!%(sm)s) { |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
142 // The normal cleanup code will take care of %(nll)s |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
143 // Py_XDECREF(%(nll)s); %(nll)s=NULL; |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
144 PyErr_SetString(PyExc_MemoryError, "failed to alloc sm output"); |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
145 %(fail)s |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
146 } |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
147 } |
30
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
148 |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
149 for (size_t i = 0; i < Nx[0]; ++i) |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
150 { |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
151 size_t j; |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
152 double sum = 0.0; |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
153 bool discount_max = false; |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
154 |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
155 const double* __restrict__ x_i = (double*)(%(x)s->data + %(x)s->strides[0] * i); |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
156 const double* __restrict__ b_i = (double*)(%(b)s->data); |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
157 const long int y_i = ((long int*)(%(y_idx)s->data + %(y_idx)s->strides[0] * i))[0]; |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
158 double* __restrict__ sm_i = (double*)(%(sm)s->data + %(sm)s->strides[0] * i); |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
159 double* __restrict__ nll_i = (double*)(%(nll)s->data + %(nll)s->strides[0] * i); |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
160 |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
161 npy_intp Sx = %(x)s->strides[1]/sizeof(double); |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
162 npy_intp Sb = %(b)s->strides[0]/sizeof(double); |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
163 npy_intp Ssm = %(sm)s->strides[1]/sizeof(double); |
24 | 164 |
30
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
165 size_t row_max_j=0; |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
166 double row_max = x_i[0] + b_i[0]; |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
167 //try to compute sum and sm the easy way |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
168 for (j = 0; j < Nx[1]; ++j) |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
169 { |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
170 double row_ij = x_i[j * Sx] + b_i[j * Sb]; |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
171 row_max_j = (row_ij > row_max) ? j : row_max_j; |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
172 row_max = (row_ij > row_max) ? row_ij : row_max; |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
173 |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
174 double sm_ij = exp(row_ij); |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
175 sum += sm_ij; |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
176 sm_i[j * Ssm] = sm_ij; |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
177 } |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
178 if ((0.0 == sum) || (isinf(sum))) |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
179 { |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
180 //our cheap trick didn't work... try again and do it better. |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
181 discount_max = true; |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
182 sum = 0.0; //reset sum and recompute.... |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
183 for (j = 0; j < Nx[1]; ++j) |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
184 { |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
185 double row_ij = x_i[j * Sx] + b_i[j * Sb]; |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
186 |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
187 double sm_ij = exp(row_ij - row_max); |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
188 sum += sm_ij; |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
189 sm_i[j * Ssm] = sm_ij; |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
190 } |
32 | 191 if ( (0.0 == sum) || (isinf(sum))) |
192 { | |
193 //that was our best... | |
194 %(fail)s; | |
195 } | |
30
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
196 //if we still can't sum it up, we're screwed. |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
197 //So far, this assertion has never failed... |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
198 } |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
199 |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
200 //cblas_dscal(x.N, 1.0 / sum, &mat_at(s,i,0), s.n); |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
201 double sum_inv = 1.0 / sum; |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
202 for (j = 0; j < Nx[1]; ++j) |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
203 { |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
204 sm_i[j * Ssm] *= sum_inv; |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
205 } |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
206 |
32 | 207 if (y_i >= Nx[1]) |
208 { | |
209 %(fail)s; | |
210 } | |
30
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
211 |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
212 nll_i[0] = - x_i[y_i*Sx] |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
213 - b_i[y_i*Sb] |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
214 + (discount_max ? row_max : 0.0) |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
215 + log(sum); |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
216 //mat_at(y,i,0) = -log( mat_at(s,i,t[i])); //less accurate? |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
217 //mat_at(y,i,0) = - mat_at(x,i,t[i]) - mat_at(b,0,t[i]) + (discount_max ? maxi : 0.0) + log(sum); |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
218 } |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
219 """ % dict(locals(), **sub) |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
220 |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
221 |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
222 |
25
b63e8c0bf21b
added __init__.py, fixed crossentropy_softmax_1hot function name
bergstrj@iro.umontreal.ca
parents:
24
diff
changeset
|
223 crossentropy_softmax_1hot = gof.op.constructor(CrossentropySoftmax1Hot) |
24 | 224 |
30
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
225 class CrossentropySoftmax1HotDx (gof.op.Op): |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
226 nin=3 |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
227 nout=1 |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
228 """Gradient wrt x of the CrossentropySoftmax1Hot Op""" |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
229 def __init__(self, dy, sm, y_idx,**kwargs): |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
230 dy = tensor._as_tensor(dy) |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
231 sm = tensor._as_tensor(sm) |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
232 y_idx = tensor._as_tensor(y_idx) |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
233 self.inputs = [dy, sm, y_idx] |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
234 self.outputs = [tensor.Tensor(sm.dtype, sm.broadcastable)] |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
235 def perform(self): |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
236 dy,sm,y_idx = [i.data for i in self.inputs] |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
237 dx = numpy.zeros_like(sm) |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
238 for i in xrange(sm.shape[0]): |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
239 dx[i] = dy[i] * sm[i] #vector scale |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
240 dx[i, y_idx[i]] -= dy[i] #scalar decrement |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
241 self.outputs[0].data = dx |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
242 def grad(self, *args): |
bf0145fa73e8
added c implementation for CrossentropySoftmax1Hot
bergstrj@iro.umontreal.ca
parents:
25
diff
changeset
|
243 raise NotImplementedError() |
32 | 244 def c_code(self, (dnll, sm, y_idx), (dx,), sub): |
245 return """ | |
246 | |
67
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
247 if ((%(dnll)s->descr->type_num != PyArray_DOUBLE) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
248 || (%(sm)s->descr->type_num != PyArray_DOUBLE) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
249 || (%(y_idx)s->descr->type_num != PyArray_INT64)) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
250 { |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
251 PyErr_SetString(PyExc_TypeError, "types should be float64, float64, int64"); |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
252 %(fail)s; |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
253 } |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
254 if ((%(dnll)s->nd != 1) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
255 || (%(sm)s->nd != 2) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
256 || (%(y_idx)s->nd != 1)) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
257 { |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
258 PyErr_SetString(PyExc_ValueError, "rank error"); |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
259 %(fail)s; |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
260 } |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
261 if ((%(dnll)s->dimensions[0] != %(sm)s->dimensions[0]) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
262 || (%(dnll)s->dimensions[0] != %(y_idx)s->dimensions[0]) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
263 || (%(dnll)s->dimensions[0] != %(dx)s->dimensions[0])) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
264 { |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
265 PyErr_SetString(PyExc_ValueError, "dimension mismatch"); |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
266 %(fail)s; |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
267 } |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
268 if ((NULL == %(dx)s) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
269 || (%(dx)s->dimensions[0] != %(sm)s->dimensions[0]) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
270 || (%(dx)s->dimensions[1] != %(sm)s->dimensions[1])) |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
271 { |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
272 if (NULL != %(dx)s) Py_XDECREF(%(dx)s); |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
273 %(dx)s = (PyArrayObject*)PyArray_SimpleNew(2, PyArray_DIMS(%(x)s), type_num_%(x)s); |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
274 if(!%(dx)s) { |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
275 // The normal cleanup code will take care of %(nll)s |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
276 // Py_XDECREF(%(nll)s); %(nll)s=NULL; |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
277 PyErr_SetString(PyExc_MemoryError, "failed to alloc dx output"); |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
278 %(fail)s |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
279 } |
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
280 } |
24 | 281 |
67
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
282 for (size_t i = 0; i < %(dx)s->dimensions[0]; ++i) |
32 | 283 { |
284 const double dnll_i = ((double*)(%(dnll)s->data + %(dnll)s->strides[0] * i))[0]; | |
285 | |
286 const long int y_i = ((long int*)(%(y_idx)s->data + %(y_idx)s->strides[0] * i))[0]; | |
287 | |
288 const double* __restrict__ sm_i = (double*)(%(sm)s->data + %(sm)s->strides[0] * i); | |
289 npy_intp Ssm = %(sm)s->strides[1]/sizeof(double); | |
290 | |
291 double* __restrict__ dx_i = (double*)(%(dx)s->data + %(dx)s->strides[0] * i); | |
292 npy_intp Sdx = %(dx)s->strides[1]/sizeof(double); | |
293 | |
67
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
294 for (size_t j = 0; j < %(dx)s->dimensions[1]; ++j) |
32 | 295 { |
296 dx_i[j * Sdx] = dnll_i * sm_i[j * Ssm]; | |
297 } | |
67
810a8e3c85e1
fixed horrible memory leak from crossentropy...
bergstra@is23.m
parents:
34
diff
changeset
|
298 if (y_i >= %(dx)s->dimensions[1]) |
32 | 299 { |
300 %(fail)s; | |
301 } | |
302 dx_i[y_i * Sdx] -= dnll_i; | |
303 } | |
304 """ % dict(locals(), **sub) |