Skip to content

P
pytensor
提交 03e77233 authored 作者: Frédéric Bastien's avatar Frédéric Bastien
浏览文件
操作

Merge pull request #3095 from harlouci/flake8_v4

flake8 for tensor/nnet/nnet.py
上级 f4edcc59 9b457370
全部展开 隐藏空白字符变更
内嵌 并排
正在显示 包含 146 行增加 和 151 行删除
theano/tensor/nnet/Conv3D.py
浏览文件 @ 03e77233
from __future__ import print_function
import numpy as N
from six.moves import xrange
import theano
from theano.tensor import basic as T
import numpy as N
#from util import strutil
# from util import strutil
from theano.tensor.blas_headers import blas_header_text, blas_header_version
from theano.tensor.blas import ldflags
from theano.misc import strutil
......@@ -72,26 +74,28 @@ class Conv3D(theano.Op):
def grad(self, inputs, output_gradients):
V, W, b, d = inputs
dCdH , = output_gradients
dCdH, = output_gradients
# make all of these ops support broadcasting of scalar b to vector b and eplace the zeros_like in all their grads
# print dCdH.broadcastable
# print "dCdH.broadcastable"
# quit(-1)
#dCdH = printing.Print("dCdH = ",["shape"])
# dCdH = printing.Print("dCdH = ",["shape"])
# Make sure the broadcasting pattern of the gradient is the the same
# as the initial variable
dCdV = ConvTransp3D.convTransp3D(W, T.zeros_like(V[0, 0, 0, 0, :]), d, dCdH, V.shape[1:4])
dCdV = theano.tensor.nnet.convTransp3D(
W, T.zeros_like(V[0, 0, 0, 0, :]), d, dCdH, V.shape[1:4])
dCdV = T.patternbroadcast(dCdV, V.broadcastable)
WShape = W.shape
dCdW = ConvGrad3D.convGrad3D(V, d, WShape, dCdH)
dCdW = theano.tensor.nnet.convGrad3D(V, d, WShape, dCdH)
dCdW = T.patternbroadcast(dCdW, W.broadcastable)
dCdb = T.sum(dCdH, axis=(0, 1, 2, 3))
dCdb = T.patternbroadcast(dCdb, b.broadcastable)
dCdd = grad_undefined(self, 3, inputs[3],
"The gradient of Conv3D with respect to the convolution" +\
" stride is undefined because Conv3D is only defined for" +\
" integer strides.")
dCdd = grad_undefined(
self, 3, inputs[3],
"The gradient of Conv3D with respect to the convolution"
" stride is undefined because Conv3D is only defined for"
" integer strides.")
if 'name' in dir(dCdH) and dCdH.name is not None:
dCdH_name = dCdH.name
......@@ -113,11 +117,13 @@ class Conv3D(theano.Op):
else:
b_name = 'anon_b'
dCdV.name = 'Conv3D_dCdV(dCdH='+dCdH_name+',V='+V_name+')'
dCdW.name = 'Conv3D_dCdW(dCdH='+dCdH_name+',V='+V_name+',W='+W_name+')'
dCdb.name = 'Conv3D_dCdb(dCdH='+dCdH_name+',V='+V_name+',W='+W_name+',b='+b_name+')'
dCdV.name = 'Conv3D_dCdV(dCdH=' + dCdH_name + ',V=' + V_name + ')'
dCdW.name = ('Conv3D_dCdW(dCdH=' + dCdH_name + ',V=' + V_name +
',W=' + W_name + ')')
dCdb.name = ('Conv3D_dCdb(dCdH=' + dCdH_name + ',V=' + V_name +
',W=' + W_name + ',b=' + b_name + ')')
return [ dCdV, dCdW, dCdb, dCdd ]
return [dCdV, dCdW, dCdb, dCdd]
def perform(self, node, inputs, output_storage):
V, W, b, d = inputs
......@@ -144,7 +150,7 @@ class Conv3D(theano.Op):
output_width = T.floor((vidWidth - filterWidth) // dc) + 1
output_dur = T.floor((vidDur - filterDur) // dt) + 1
rval = (batch_size, output_height, output_width, output_dur, output_channels )
rval = (batch_size, output_height, output_width, output_dur, output_channels)
return [rval]
......@@ -155,7 +161,7 @@ class Conv3D(theano.Op):
return ldflags()
def c_compile_args(self):
flags = ldflags(libs=False, flags=True)
flags = ldflags(libs=False, flags=True)
return flags
def c_lib_dirs(self):
......@@ -170,7 +176,7 @@ class Conv3D(theano.Op):
H = outputs[0]
codeSource = """
codeSource = """
///////////// < code generated by Conv3D >
//printf("\t\t\t\tConv3D c code\\n");
......@@ -320,13 +326,13 @@ class Conv3D(theano.Op):
VV, WV, bv, dv = node.inputs
HV = node.outputs[0]
if (theano.config.blas.ldflags and
VV.dtype == WV.dtype and HV.dtype == VV.dtype):
VV.dtype == WV.dtype and HV.dtype == VV.dtype):
if VV.dtype == 'float64':
gemv = 'dgemv_'
elif VV.dtype == 'float32':
gemv = 'sgemv_'
else:
raise Exception('Unrecognized dtype for convolution '+V.value.dtype)
raise Exception('Unrecognized dtype for convolution ' + V.value.dtype)
codeSource += """
if (inputChannels > 20 && outputChannels > 20 && ws4 == sizeof(ELEM_AT(%(W)s,0)))
......@@ -571,7 +577,7 @@ def computeH(V, W, b, d):
outputChannels = W.shape[0]
inputChannels = V.shape[4]
if W.shape[4] != inputChannels:
raise Exception("W.shape[4] = "+str(W.shape[4])+" but inputChannels = "+str(inputChannels))
raise Exception("W.shape[4] = " + str(W.shape[4]) + " but inputChannels = " + str(inputChannels))
filterHeight = W.shape[1]
filterWidth = W.shape[2]
filterDur = W.shape[3]
......@@ -586,12 +592,12 @@ def computeH(V, W, b, d):
assert dy > 0
assert dt > 0
outputHeight = int( (vidHeight - filterHeight) / dx )+1
outputWidth = int( (vidWidth - filterWidth) / dy )+1
outputDur = int( (vidDur - filterDur) / dt ) + 1
outputHeight = int((vidHeight - filterHeight) / dx) + 1
outputWidth = int((vidWidth - filterWidth) / dy) + 1
outputDur = int((vidDur - filterDur) / dt) + 1
H = N.zeros( (batchSize, outputHeight,
outputWidth, outputDur, outputChannels ), dtype=V.dtype )
H = N.zeros((batchSize, outputHeight,
outputWidth, outputDur, outputChannels), dtype=V.dtype)
# H[i,j,x,y,t] = b_j + sum_k sum_l sum_m sum_z W[j,z,k,l,m] V[i,z, dx*x+k,dy*y+l,dt*t+m]
for i in xrange(0, H.shape[0]):
......@@ -610,12 +616,8 @@ def computeH(V, W, b, d):
# if (i,j,x,y,t) == (0,0,0,0,0):
# print (( W[j,z,k,l,m] , V[i,z,d[0]*x+k,d[1]*y+l,d[2]*t+m] ), (k,l,m) )
w = W[j, k, l, m, z]
v = V[i, d[0]*x+k, d[1]*y+l, d[2]*t+m, z]
v = V[i, d[0] * x + k, d[1] * y + l, d[2] * t + m, z]
# if i == 0 and x == 0 and y == 0 and t == 0 and j == 0:
# print 'setting H[0] += '+str(w*v)+' W['+str((j,z,k,l,m))+']='+str(w)+' V['+str((i,d[0]*x+k,d[1]*y+l,d[2]*t+m,z))+']='+str(v)
H[i, x, y, t, j] += w * v
return H
from . import ConvGrad3D
from . import ConvTransp3D
theano/tensor/nnet/ConvGrad3D.py
浏览文件 @ 03e77233
from six.moves import xrange
import numpy as N
import theano
from theano.tensor import basic as T
from theano.misc import strutil
import numpy as N
from six.moves import xrange
from theano.gradient import grad_undefined
from theano.gradient import DisconnectedType
......@@ -23,11 +25,15 @@ class ConvGrad3D(theano.Op):
WShape_ = T.as_tensor_variable(WShape)
dCdH_ = T.as_tensor_variable(dCdH)
return theano.Apply(self, inputs=[V_, d_, WShape_, dCdH_], outputs=[ T.TensorType(V_.dtype, (False, False, False, False, False))() ] )
return theano.Apply(self,
inputs=[V_, d_, WShape_, dCdH_],
outputs=[T.TensorType(
V_.dtype,
(False, False, False, False, False))()])
def infer_shape(self, node, input_shapes):
V, d, W_shape, dCdH = node.inputs
return [ ( W_shape[0], W_shape[1], W_shape[2], W_shape[3], W_shape[4] ) ]
return [(W_shape[0], W_shape[1], W_shape[2], W_shape[3], W_shape[4])]
def connection_pattern(self, node):
......@@ -38,12 +44,12 @@ class ConvGrad3D(theano.Op):
dLdA, = output_gradients
z = T.zeros_like(C[0, 0, 0, 0, :])
dLdC = convTransp3D(dLdA, z, d, B, C.shape[1:4])
dLdC = theano.tensor.nnet.convTransp3D(dLdA, z, d, B, C.shape[1:4])
# d actually does affect the outputs, so it's not disconnected
dLdd = grad_undefined(self, 1, d)
# The shape of the weights doesn't affect the output elements
dLdWShape = DisconnectedType()()
dLdB = conv3D(C, dLdA, T.zeros_like(B[0, 0, 0, 0, :]), d)
dLdB = theano.tensor.nnet.conv3D(C, dLdA, T.zeros_like(B[0, 0, 0, 0, :]), d)
return [dLdC, dLdd, dLdWShape, dLdB]
......@@ -54,15 +60,10 @@ class ConvGrad3D(theano.Op):
# partial C / partial W[j,z,k,l,m] = sum_i sum_p sum_q sum_r (partial C /partial H[i,j,p,q,r] ) * V[i,z,dr*p+k,dc*q+l,dt*r+m]
batchSize = dCdH.shape[0]
outputFilters = dCdH.shape[4]
outputHeight = dCdH.shape[1]
outputWidth = dCdH.shape[2]
outputDur = dCdH.shape[3]
assert V.shape[0] == batchSize
inputFilters = V.shape[4]
inputHeight = V.shape[1]
inputWidth = V.shape[2]
inputDur = V.shape[3]
dr, dc, dt = d
dCdW = N.zeros(WShape, dtype=V.dtype)
......@@ -78,7 +79,10 @@ class ConvGrad3D(theano.Op):
for r in xrange(0, outputDur):
for j in xrange(0, WShape[0]):
for z in xrange(0, WShape[4]):
dCdW[j, k, l, m, z] += dCdH[i, p, q, r, j] * V[i, dr*p+k, dc*q+l, dt*r+m, z]
dCdW[j, k, l, m, z] += (
dCdH[i, p, q, r, j] *
V[i, dr * p + k, dc * q + l,
dt * r + m, z])
output_storage[0][0] = dCdW
......@@ -272,6 +276,3 @@ class ConvGrad3D(theano.Op):
convGrad3D = ConvGrad3D()
from theano.tensor.nnet.Conv3D import conv3D
from theano.tensor.nnet.ConvTransp3D import convTransp3D
theano/tensor/nnet/ConvTransp3D.py
浏览文件 @ 03e77233
from __future__ import print_function
import numpy as N
from six.moves import xrange
import theano
from theano.tensor import basic as T
from theano.misc import strutil
import theano
from theano.gradient import grad_undefined
from theano.gradient import DisconnectedType
......@@ -31,12 +33,15 @@ class ConvTransp3D(theano.Op):
else:
RShape_ = T.as_tensor_variable([-1, -1, -1])
return theano.Apply(self, inputs=[W_, b_, d_, H_, RShape_], outputs=[ T.TensorType(H_.dtype, (False, False, False, False, False))() ] )
return theano.Apply(self,
inputs=[W_, b_, d_, H_, RShape_],
outputs=[T.TensorType(H_.dtype,
(False, False, False, False, False))()])
def infer_shape(self, node, input_shapes):
W, b, d, H, RShape = node.inputs
W_shape, b_shape, d_shape, H_shape, RShape_shape = input_shapes
return [(H_shape[0], RShape[0], RShape[1], RShape[2], W_shape[4])]
return [(H_shape[0], RShape[0], RShape[1], RShape[2], W_shape[4])]
def connection_pattern(self, node):
return [[True], [True], [True], [True], [False]]
......@@ -44,9 +49,9 @@ class ConvTransp3D(theano.Op):
def grad(self, inputs, output_gradients):
W, b, d, H, RShape = inputs
dCdR, = output_gradients
dCdH = conv3D(dCdR, W, T.zeros_like(H[0, 0, 0, 0, :]), d)
dCdH = theano.tensor.nnet.conv3D(dCdR, W, T.zeros_like(H[0, 0, 0, 0, :]), d)
WShape = W.shape
dCdW = convGrad3D(dCdR, d, WShape, H)
dCdW = theano.tensor.nnet.convGrad3D(dCdR, d, WShape, H)
dCdb = T.sum(dCdR, axis=(0, 1, 2, 3))
# not differentiable, since d affects the output elements
dCdd = grad_undefined(self, 2, d)
......@@ -73,11 +78,13 @@ class ConvTransp3D(theano.Op):
else:
b_name = 'anon_b'
dCdW.name = 'ConvTransp3D_dCdW.H='+H_name+',dCdR='+dCdR_name+',W='+W_name
dCdb.name = 'ConvTransp3D_dCdb.H='+H_name+',dCdR='+dCdR_name+',W='+W_name+',b='+b_name
dCdW.name = ('ConvTransp3D_dCdW.H=' + H_name + ',dCdR=' + dCdR_name +
',W=' + W_name)
dCdb.name = ('ConvTransp3D_dCdb.H=' + H_name + ',dCdR=' + dCdR_name +
',W=' + W_name + ',b=' + b_name)
dCdH.name = 'ConvTransp3D_dCdH.H=' + H_name + ',dCdR=' + dCdR_name
return [dCdW, dCdb, dCdd, dCdH, dCdRShape]
return [dCdW, dCdb, dCdd, dCdH, dCdRShape]
def perform(self, node, inputs, output_storage):
W, b, d, H, RShape = inputs
......@@ -335,7 +342,7 @@ def computeR(W, b, d, H, Rshape=None):
assert len(b.shape) == 1
assert len(d) == 3
outputChannels, filterHeight, filterWidth, filterDur, \
outputChannels, filterHeight, filterWidth, filterDur, \
inputChannels = W.shape
batchSize, outputHeight, outputWidth, outputDur, \
outputChannelsAgain = H.shape
......@@ -367,7 +374,7 @@ def computeR(W, b, d, H, Rshape=None):
# print "video size: "+str((videoHeight, videoWidth, videoDur))
R = N.zeros((batchSize, videoHeight,
videoWidth, videoDur, inputChannels), dtype=H.dtype)
videoWidth, videoDur, inputChannels), dtype=H.dtype)
# R[i,j,r,c,t] = b_j + sum_{rc,rk | d \circ rc + rk = r} sum_{cc,ck | ...} sum_{tc,tk | ...} sum_k W[k, j, rk, ck, tk] * H[i,k,rc,cc,tc]
for i in xrange(0, batchSize):
......@@ -404,8 +411,8 @@ def computeR(W, b, d, H, Rshape=None):
if tk < 0:
break
R[
i, r, c, t, j] += N.dot(W[:, rk, ck, tk, j], H[i, rc, cc, tc, :] )
R[i, r, c, t, j] += N.dot(
W[:, rk, ck, tk, j], H[i, rc, cc, tc, :])
tc += 1
"" # close loop over tc
......@@ -421,7 +428,3 @@ def computeR(W, b, d, H, Rshape=None):
"" # close loop over i
return R
from theano.tensor.nnet.Conv3D import conv3D
from theano.tensor.nnet.ConvGrad3D import convGrad3D
theano/tensor/nnet/conv.py
浏览文件 @ 03e77233
差异被折叠。
theano/tensor/nnet/conv3d2d.py
浏览文件 @ 03e77233
......@@ -194,13 +194,13 @@ def conv3d(signals, filters,
_signals_shape_5d[2],
_signals_shape_5d[3],
_signals_shape_5d[4],
)
)
_filters_shape_4d = (
_filters_shape_5d[0] * _filters_shape_5d[1],
_filters_shape_5d[2],
_filters_shape_5d[3],
_filters_shape_5d[4],
)
)
if border_mode[1] != border_mode[2]:
raise NotImplementedError('height and width bordermodes must match')
......@@ -228,7 +228,7 @@ def conv3d(signals, filters,
_filters_shape_5d[1], # Tf
_signals_shape_5d[3] - _filters_shape_5d[3] + 1,
_signals_shape_5d[4] - _filters_shape_5d[4] + 1,
))
))
elif border_mode[1] == 'full':
out_tmp = out_4d.reshape((
_signals_shape_5d[0], # Ns
......@@ -237,7 +237,7 @@ def conv3d(signals, filters,
_filters_shape_5d[1], # Tf
_signals_shape_5d[3] + _filters_shape_5d[3] - 1,
_signals_shape_5d[4] + _filters_shape_5d[4] - 1,
))
))
elif border_mode[1] == 'same':
raise NotImplementedError()
else:
......@@ -246,15 +246,15 @@ def conv3d(signals, filters,
# now sum out along the Tf to get the output
# but we have to sum on a diagonal through the Tf and Ts submatrix.
if border_mode[0] == 'valid':
if _filters_shape_5d[1]!=1:
out_5d = diagonal_subtensor(out_tmp, 1, 3).sum(axis=3)
else: # for Tf==1, no sum along Tf, the Ts-axis of the output is unchanged!
out_5d = out_tmp.reshape((
_signals_shape_5d[0],
_signals_shape_5d[1],
_filters_shape_5d[0],
_signals_shape_5d[3] - _filters_shape_5d[3] + 1,
_signals_shape_5d[4] - _filters_shape_5d[4] + 1,
if _filters_shape_5d[1] != 1:
out_5d = diagonal_subtensor(out_tmp, 1, 3).sum(axis=3)
else: # for Tf==1, no sum along Tf, the Ts-axis of the output is unchanged!
out_5d = out_tmp.reshape((
_signals_shape_5d[0],
_signals_shape_5d[1],
_filters_shape_5d[0],
_signals_shape_5d[3] - _filters_shape_5d[3] + 1,
_signals_shape_5d[4] - _filters_shape_5d[4] + 1,
))
elif border_mode[0] in ('full', 'same'):
raise NotImplementedError('sequence border mode', border_mode[0])
......@@ -316,7 +316,7 @@ if cuda.cuda_available:
def local_inplace_DiagonalSubtensor(node):
""" also work for IncDiagonalSubtensor """
if (isinstance(node.op, (DiagonalSubtensor, IncDiagonalSubtensor)) and
not node.op.inplace):
not node.op.inplace):
new_op = node.op.__class__(inplace=True)
new_node = new_op(*node.inputs)
return [new_node]
......
theano/tensor/nnet/neighbours.py
浏览文件 @ 03e77233
......@@ -2,15 +2,15 @@
TODO: implement Images2Neibs.infer_shape() methods
"""
from six.moves import xrange
import numpy
import theano
from theano import Op, Apply
import theano.tensor as T
from theano.gradient import grad_not_implemented
from theano.gradient import grad_undefined
import numpy
class Images2Neibs(Op):
......@@ -206,7 +206,7 @@ class Images2Neibs(Op):
z_col = j + d * i
z[0][z_row, z_col] = ten4[n, s, ten4_2, ten4_3]
def infer_shape(self, node, input_shape):
in_shape = input_shape[0]
c, d = node.inputs[1]
......@@ -223,7 +223,7 @@ class Images2Neibs(Op):
z_dim0 = grid_c * grid_d * in_shape[1] * in_shape[0]
z_dim1 = c * d
return [(z_dim0, z_dim1)]
def c_code(self, node, name, inp, out, sub):
ten4, neib_shape, neib_step = inp
z, = out
......@@ -417,21 +417,21 @@ class Images2Neibs(Op):
def images2neibs(ten4, neib_shape, neib_step=None, mode='valid'):
"""
"""
Function :func:`images2neibs <theano.sandbox.neighbours.images2neibs>`
allows to apply a sliding window operation to a tensor containing
allows to apply a sliding window operation to a tensor containing
images
or other two-dimensional objects.
The sliding window operation loops
over points in input data and stores a rectangular neighbourhood of
each point.
It is possible to assign a step of selecting patches (parameter
`neib_step`).
:param ten4: A 4-dimensional tensor which represents
or other two-dimensional objects.
The sliding window operation loops
over points in input data and stores a rectangular neighbourhood of
each point.
It is possible to assign a step of selecting patches (parameter
`neib_step`).
:param ten4: A 4-dimensional tensor which represents
a list of lists of images.a list of lists of images.
It should have shape (list 1 dim, list 2 dim,
row, col). The first two dimensions can be
row, col). The first two dimensions can be
useful to store different channels and batches.
:type ten4: A 4d tensor-like.
:param neib_shape: A tuple containing two
......@@ -442,20 +442,20 @@ def images2neibs(ten4, neib_shape, neib_step=None, mode='valid'):
:type neib_shape: A 1d tensor-like of 2 values.
:param neib_step: (dr,dc) where dr is the number of rows to
skip between patch and dc is the number of
columns. The parameter should be a tuple of two elements:
number
of rows and number of columns to skip each iteration.
columns. The parameter should be a tuple of two elements:
number
of rows and number of columns to skip each iteration.
Basically, when the step is 1, the neighbourhood of every
first element is taken and every possible rectangular
first element is taken and every possible rectangular
subset is returned. By default it is equal to
`neib_shape` in other words, the
patches are disjoint. When the step is greater than
patches are disjoint. When the step is greater than
`neib_shape`, some elements are omitted. When None, this
is the same as
neib_shape(patch are disjoint)
.. note:: Currently the step size should be chosen in the way that the
corresponding dimension :math:`i` (width or height) is equal to
.. note:: Currently the step size should be chosen in the way that the
corresponding dimension :math:`i` (width or height) is equal to
:math:`n * step\_size_i + neib\_shape_i` for some :math:`n`
:type neib_step: A 1d tensor-like of 2 values.
:param mode:
......@@ -489,29 +489,29 @@ def images2neibs(ten4, neib_shape, neib_step=None, mode='valid'):
= flattened version of ten4[i,j,l:l+r,k:k+c]
idx += 1
.. note:: The operation isn't necessarily implemented internally with
these for loops, they're just the easiest way to describe the
.. note:: The operation isn't necessarily implemented internally with
these for loops, they're just the easiest way to describe the
output pattern.
Example:
.. code-block:: python
# Defining variables
images = T.tensor4('images')
neibs = images2neibs(images, neib_shape=(5, 5))
# Constructing theano function
# Constructing theano function
window_function = theano.function([images], neibs)
# Input tensor (one image 10x10)
im_val = np.arange(100.).reshape((1, 1, 10, 10))
# Function application
neibs_val = window_function(im_val)
.. note:: The underlying code will construct a 2D tensor of disjoint
patches 5x5. The output has shape 4x25.
.. note:: The underlying code will construct a 2D tensor of disjoint
patches 5x5. The output has shape 4x25.
"""
return Images2Neibs(mode)(ten4, neib_shape, neib_step)
......@@ -524,25 +524,24 @@ def neibs2images(neibs, neib_shape, original_shape, mode='valid'):
the output of :func:`images2neibs <theano.sandbox.neigbours.neibs2images>`
and reconstructs its input.
:param neibs: matrix like the one obtained by
:param neibs: matrix like the one obtained by
:func:`images2neibs <theano.sandbox.neigbours.neibs2images>`
:param neib_shape: `neib_shape` that was used in
:param neib_shape: `neib_shape` that was used in
:func:`images2neibs <theano.sandbox.neigbours.neibs2images>`
:param original_shape: original shape of the 4d tensor given to
:param original_shape: original shape of the 4d tensor given to
:func:`images2neibs <theano.sandbox.neigbours.neibs2images>`
:return: Reconstructs the input of
:return: Reconstructs the input of
:func:`images2neibs <theano.sandbox.neigbours.neibs2images>`,
a 4d tensor of shape `original_shape`.
.. note:: Currently, the function doesn't support tensors created with
`neib_step` different from default value. This means that it may be
impossible to compute the gradient of a variable gained by
:func:`images2neibs <theano.sandbox.neigbours.neibs2images>` w.r.t.
its inputs in this case, because it uses
:func:`images2neibs <theano.sandbox.neigbours.neibs2images>` for
impossible to compute the gradient of a variable gained by
:func:`images2neibs <theano.sandbox.neigbours.neibs2images>` w.r.t.
its inputs in this case, because it uses
:func:`images2neibs <theano.sandbox.neigbours.neibs2images>` for
gradient computation.
Example, which uses a tensor gained in example for
:func:`images2neibs <theano.sandbox.neigbours.neibs2images>`:
......
theano/tensor/nnet/nnet.py
浏览文件 @ 03e77233
差异被折叠。
theano/tensor/nnet/sigm.py
浏览文件 @ 03e77233
......@@ -7,7 +7,6 @@ from __future__ import print_function
import warnings
import numpy
from six.moves import xrange
import theano
from theano import config, gof, printing, scalar
......@@ -92,7 +91,7 @@ class ScalarSigmoid(scalar.UnaryScalarOp):
x, = inp
z, = out
if (not theano.config.lib.amdlibm or
node.inputs[0].dtype != node.outputs[0].dtype):
node.inputs[0].dtype != node.outputs[0].dtype):
raise theano.gof.utils.MethodNotDefined()
dtype = node.inputs[0].dtype
if dtype == 'float32' and self.amd_float32 is not None:
......@@ -129,9 +128,8 @@ class ScalarSigmoid(scalar.UnaryScalarOp):
"""
This method was used to generate the graph: sigmoid_prec.png in the doc
"""
import matplotlib
data = numpy.arange(-15, 15, .1)
val = 1/(1+numpy.exp(-data))
val = 1 / (1 + numpy.exp(-data))
def hard_sigmoid(x):
return theano.tensor.nnet.hard_sigmoid(x)
......@@ -164,10 +162,10 @@ scalar_sigmoid = ScalarSigmoid(scalar.upgrade_to_float, name='scalar_sigmoid')
sigmoid = elemwise.Elemwise(scalar_sigmoid, name='sigmoid')
sigmoid_inplace = elemwise.Elemwise(
ScalarSigmoid(scalar.transfer_type(0)),
inplace_pattern={0: 0},
name='sigmoid_inplace',
)
ScalarSigmoid(scalar.transfer_type(0)),
inplace_pattern={0: 0},
name='sigmoid_inplace',
)
pprint.assign(sigmoid, printing.FunctionPrinter('sigmoid'))
......@@ -240,7 +238,7 @@ pprint.assign(ultra_fast_sigmoid,
printing.FunctionPrinter('ultra_fast_sigmoid'))
#@opt.register_uncanonicalize
# @opt.register_uncanonicalize
@gof.local_optimizer([sigmoid])
def local_ultra_fast_sigmoid(node):
"""
......@@ -290,7 +288,7 @@ def hard_sigmoid(x):
return x
#@opt.register_uncanonicalize
# @opt.register_uncanonicalize
@gof.local_optimizer([sigmoid])
def local_hard_sigmoid(node):
if (isinstance(node.op, tensor.Elemwise) and
......@@ -412,7 +410,7 @@ def is_1pexp(t):
"""
if t.owner and t.owner.op == tensor.add:
scalars, scalar_inputs, nonconsts = \
opt.scalarconsts_rest(t.owner.inputs)
opt.scalarconsts_rest(t.owner.inputs)
# scalar_inputs are potentially dimshuffled and fill'd scalars
if len(nonconsts) == 1:
maybe_exp = nonconsts[0]
......@@ -439,11 +437,12 @@ def is_1pexp(t):
return None
AddConfigVar('warn.identify_1pexp_bug',
'Warn if Theano versions prior to 7987b51 (2011-12-18) could have '
'yielded a wrong result due to a bug in the is_1pexp function',
BoolParam(theano.configdefaults.warn_default('0.4.1')),
in_c_key=False)
AddConfigVar(
'warn.identify_1pexp_bug',
'Warn if Theano versions prior to 7987b51 (2011-12-18) could have '
'yielded a wrong result due to a bug in the is_1pexp function',
BoolParam(theano.configdefaults.warn_default('0.4.1')),
in_c_key=False)
def is_exp(var):
......@@ -778,9 +777,9 @@ def perform_sigm_times_exp(tree, exp_x=None, exp_minus_x=None, sigm_x=None,
rval = False
for sub_idx, sub_tree in enumerate(inputs):
rval |= perform_sigm_times_exp(
tree=sub_tree, parent=tree, child_idx=sub_idx,
exp_x=exp_x, exp_minus_x=exp_minus_x, sigm_x=sigm_x,
sigm_minus_x=sigm_minus_x, full_tree=full_tree)
tree=sub_tree, parent=tree, child_idx=sub_idx,
exp_x=exp_x, exp_minus_x=exp_minus_x, sigm_x=sigm_x,
sigm_minus_x=sigm_minus_x, full_tree=full_tree)
return rval
else:
# Reached a leaf: if it is an exponential or a sigmoid, then we
......@@ -867,15 +866,15 @@ def local_inv_1_plus_exp(node):
inv_arg = node.inputs[0]
if inv_arg.owner and inv_arg.owner.op == tensor.add:
scalars, scalar_inputs, nonconsts = \
opt.scalarconsts_rest(inv_arg.owner.inputs)
opt.scalarconsts_rest(inv_arg.owner.inputs)
# scalar_inputs are potentially dimshuffled and fill'd scalars
if len(nonconsts) == 1:
if nonconsts[0].owner and nonconsts[0].owner.op == tensor.exp:
if scalars and numpy.allclose(numpy.sum(scalars), 1):
return opt._fill_chain(
sigmoid(
tensor.neg(nonconsts[0].owner.inputs[0])),
scalar_inputs)
sigmoid(
tensor.neg(nonconsts[0].owner.inputs[0])),
scalar_inputs)
# Registration is below, and conditional.
......@@ -892,7 +891,7 @@ def local_1msigmoid(node):
if sub_r.owner and sub_r.owner.op == sigmoid:
try:
val_l = opt.get_scalar_constant_value(sub_l)
except Exception as e:
except Exception:
return
if numpy.allclose(numpy.sum(val_l), 1):
return [sigmoid(-sub_r.owner.inputs[0])]
......@@ -921,7 +920,6 @@ if 0:
print(sigm_canonicalize(node))
def sigm_canonicalize(node):
add = tensor.add
mul = tensor.mul
div = tensor.true_div
......
theano/tests/test_flake8.py
浏览文件 @ 03e77233
......@@ -88,15 +88,7 @@ whitelist_flake8 = [
"tensor/signal/conv.py",
"tensor/signal/tests/test_conv.py",
"tensor/signal/tests/test_downsample.py",
"tensor/nnet/nnet.py",
"tensor/nnet/Conv3D.py",
"tensor/nnet/__init__.py",
"tensor/nnet/ConvTransp3D.py",
"tensor/nnet/sigm.py",
"tensor/nnet/ConvGrad3D.py",
"tensor/nnet/conv3d2d.py",
"tensor/nnet/conv.py",
"tensor/nnet/neighbours.py",
"tensor/nnet/tests/test_conv.py",
"tensor/nnet/tests/test_neighbours.py",
"tensor/nnet/tests/test_nnet.py",
......
Markdown 格式
0% 或
您添加了 0 人 到此讨论。请谨慎行事。
请先完成此评论的编辑！
请 注册 或者 后发表评论