Skip to content

P
pytensor
提交 791d4871 authored 作者: Chiheb Trabelsi's avatar Chiheb Trabelsi
浏览文件
操作

test_conv_cuda_ndarray.py has been modified in order to respect the flake8 style.

上级 186b90a0
显示空白字符变更
内嵌 并排
正在显示 包含 110 行增加 和 114 行删除
theano/sandbox/cuda/tests/test_conv_cuda_ndarray.py
浏览文件 @ 791d4871
...@@ -2,14 +2,16 @@ ...@@ -2,14 +2,16 @@
Tests for GPU convolution Tests for GPU convolution
""" """
from __future__ import absolute_import, print_function, division from __future__ import absolute_import, print_function, division
import sys
import time import time
import unittest import unittest
import traceback import theano
from theano import tensor
from theano.tests.unittest_tools import seed_rng, assert_allclose
from theano.sandbox import cuda
import numpy import numpy
from six.moves import xrange from six.moves import xrange
from theano.sandbox.cuda.dnn import GpuDnnConv, DnnBase, dnn_conv
from nose.plugins.skip import SkipTest from nose.plugins.skip import SkipTest
from nose.tools import assert_raises from nose.tools import assert_raises
imported_scipy_convolve2d = False imported_scipy_convolve2d = False
...@@ -19,16 +21,10 @@ try: ...@@ -19,16 +21,10 @@ try:
except ImportError: except ImportError:
pass pass
import theano
from theano import tensor
from theano.tests.unittest_tools import seed_rng, assert_allclose
# Skip test if cuda is not available. # Skip test if cuda is not available.
from theano.sandbox import cuda if cuda.cuda_available is False:
if cuda.cuda_available == False:
raise SkipTest('Optional package cuda disabled') raise SkipTest('Optional package cuda disabled')
from theano.sandbox.cuda.dnn import GpuDnnConv, DnnBase, dnn_conv
# needed as the gpu conv don't have a perform implementation. # needed as the gpu conv don't have a perform implementation.
if theano.config.mode == 'FAST_COMPILE': if theano.config.mode == 'FAST_COMPILE':
...@@ -106,11 +102,11 @@ def py_conv(img, kern, mode, subsample): ...@@ -106,11 +102,11 @@ def py_conv(img, kern, mode, subsample):
if imported_scipy_convolve2d: if imported_scipy_convolve2d:
return py_conv_scipy(img, kern, mode, subsample) return py_conv_scipy(img, kern, mode, subsample)
elif mode == 'valid': elif mode == 'valid':
return py_conv_valid_numpy(img, kern)[:, :, ::subsample[0], return py_conv_valid_numpy(img, kern)[
::subsample[1]] :, :, ::subsample[0], ::subsample[1]]
elif mode == 'full': elif mode == 'full':
return py_conv_full_numpy(img, kern)[:, :, ::subsample[0], return py_conv_full_numpy(img, kern)[
::subsample[1]] :, :, ::subsample[0], ::subsample[1]]
else: else:
raise Exception("Can't execute this kernel.") raise Exception("Can't execute this kernel.")
...@@ -129,7 +125,7 @@ def py_conv_scipy(img, kern, mode, subsample): ...@@ -129,7 +125,7 @@ def py_conv_scipy(img, kern, mode, subsample):
for b in xrange(out.shape[0]): for b in xrange(out.shape[0]):
for k in xrange(out.shape[1]): for k in xrange(out.shape[1]):
for s in xrange(img.shape[1]): for s in xrange(img.shape[1]):
#convolve2d or correlate # convolve2d or correlate
out[b, k, :, :] += convolve2d(img[b, s, :, :], out[b, k, :, :] += convolve2d(img[b, s, :, :],
kern[k, s, :, :], kern[k, s, :, :],
mode) mode)
...@@ -168,10 +164,12 @@ def _params_allgood(ishape, kshape, mode, subsample=(1, 1), img_stride=(1, 1), ...@@ -168,10 +164,12 @@ def _params_allgood(ishape, kshape, mode, subsample=(1, 1), img_stride=(1, 1),
npy_kern = theano._asarray(numpy.random.rand(*kshape) - 2, npy_kern = theano._asarray(numpy.random.rand(*kshape) - 2,
dtype='float32') dtype='float32')
else: else:
npy_img = theano._asarray(numpy.arange( npy_img = theano._asarray(
numpy.prod(ishape)).reshape(ishape), dtype='float32') + 1 numpy.arange(numpy.prod(ishape)).reshape(ishape),
npy_kern = -(theano._asarray(numpy.arange( dtype='float32') + 1
numpy.prod(kshape)).reshape(kshape), dtype='float32') + 1) npy_kern = -(theano._asarray(
numpy.arange(numpy.prod(kshape)).reshape(kshape),
dtype='float32') + 1)
img = cuda_ndarray.CudaNdarray(npy_img) img = cuda_ndarray.CudaNdarray(npy_img)
kern = cuda_ndarray.CudaNdarray(npy_kern) kern = cuda_ndarray.CudaNdarray(npy_kern)
...@@ -281,15 +279,15 @@ def get_shapes(imshp=(1, 1), kshp=(1, 1), subsample=(1, 1), ...@@ -281,15 +279,15 @@ def get_shapes(imshp=(1, 1), kshp=(1, 1), subsample=(1, 1),
((3, 1) + imshp, (1, 1) + kshp, subsample, img_stride, kern_stride), ((3, 1) + imshp, (1, 1) + kshp, subsample, img_stride, kern_stride),
# nkern only # nkern only
((1, 1) + imshp, (2, 1) + kshp, subsample, img_stride, kern_stride), ((1, 1) + imshp, (2, 1) + kshp, subsample, img_stride, kern_stride),
#batch and nkern # batch and nkern
((3, 1) + imshp, (2, 1) + kshp, subsample, img_stride, kern_stride), ((3, 1) + imshp, (2, 1) + kshp, subsample, img_stride, kern_stride),
#batch and stack # batch and stack
((3, 2) + imshp, (1, 2) + kshp, subsample, img_stride, kern_stride), ((3, 2) + imshp, (1, 2) + kshp, subsample, img_stride, kern_stride),
#stack and nkern # stack and nkern
((1, 2) + imshp, (2, 2) + kshp, subsample, img_stride, kern_stride), ((1, 2) + imshp, (2, 2) + kshp, subsample, img_stride, kern_stride),
#batch, nkern and stack # batch, nkern and stack
((2, 2) + imshp, (2, 2) + kshp, subsample, img_stride, kern_stride), ((2, 2) + imshp, (2, 2) + kshp, subsample, img_stride, kern_stride),
#batch, nkern and stack # batch, nkern and stack
((3, 2) + imshp, (4, 2) + kshp, subsample, img_stride, kern_stride) ((3, 2) + imshp, (4, 2) + kshp, subsample, img_stride, kern_stride)
] ]
...@@ -345,37 +343,37 @@ def get_valid_shapes(): ...@@ -345,37 +343,37 @@ def get_valid_shapes():
shapes += [ shapes += [
# other test # other test
((2, 1, 2, 2), (1, 1, 2, 2), (1, 1), (1, 1), (1, 1)) ((2, 1, 2, 2), (1, 1, 2, 2), (1, 1), (1, 1), (1, 1)),
, ((3, 2, 4, 4), (4, 2, 4, 4), (1, 1), (1, 1), (1, 1)) ((3, 2, 4, 4), (4, 2, 4, 4), (1, 1), (1, 1), (1, 1)),
, ((4, 1, 10, 10), (1, 1, 2, 2), (1, 1), (1, 1), (1, 1)) ((4, 1, 10, 10), (1, 1, 2, 2), (1, 1), (1, 1), (1, 1)),
, ((1, 1, 4, 4), (1, 1, 2, 3), (1, 1), (1, 1), (1, 1)) ((1, 1, 4, 4), (1, 1, 2, 3), (1, 1), (1, 1), (1, 1)),
, ((4, 1, 10, 10), (1, 1, 2, 3), (1, 1), (1, 1), (1, 1)) ((4, 1, 10, 10), (1, 1, 2, 3), (1, 1), (1, 1), (1, 1)),
, ((4, 1, 10, 10), (1, 1, 2, 10), (1, 1), (1, 1), (1, 1)) ((4, 1, 10, 10), (1, 1, 2, 10), (1, 1), (1, 1), (1, 1)),
, ((4, 1, 20, 10), (1, 1, 2, 10), (1, 1), (1, 1), (1, 1)) ((4, 1, 20, 10), (1, 1, 2, 10), (1, 1), (1, 1), (1, 1)),
, ((3, 2, 8, 8), (4, 2, 4, 4), (1, 1), (1, 1), (1, 1)) # stack, nkern, bsize ((3, 2, 8, 8), (4, 2, 4, 4), (1, 1), (1, 1), (1, 1)), # stack, nkern, bsize,
, ((3, 2, 8, 6), (4, 2, 4, 4), (1, 1), (1, 1), (1, 1)) # stack, nkern, bsize, non-square image ((3, 2, 8, 6), (4, 2, 4, 4), (1, 1), (1, 1), (1, 1)), # stack, nkern, bsize, non-square image,
, ((3, 2, 8, 6), (4, 2, 4, 3), (1, 1), (1, 1), (1, 1)) # stack, nkern, bsize, non-square image, non-square kern ((3, 2, 8, 6), (4, 2, 4, 3), (1, 1), (1, 1), (1, 1)), # stack, nkern, bsize, non-square image, non-square kern,
, ((3, 2, 8, 6), (4, 2, 4, 6), (1, 1), (1, 1), (1, 1)) # stack, nkern, bsize ,non-square image, non-square kern, kernsize==imgsize on one dim ((3, 2, 8, 6), (4, 2, 4, 6), (1, 1), (1, 1), (1, 1)), # stack, nkern, bsize ,non-square image, non-square kern, kernsize==imgsize on one dim,
, ((16, 5, 64, 64), (8, 5, 8, 8), (1, 1), (1, 1), (1, 1)) # a big one ((16, 5, 64, 64), (8, 5, 8, 8), (1, 1), (1, 1), (1, 1)), # a big one
, ((16, 1, 28, 28), (20, 1, 5, 5), (1, 1), (1, 1), (1, 1)) # MNIST LeNET layer 1 ((16, 1, 28, 28), (20, 1, 5, 5), (1, 1), (1, 1), (1, 1)), # MNIST LeNET layer 1
, ((20, 16, 32, 32), (1, 16, 28, 28), (1, 1), (1, 1), (1, 1)) # layer 1 backprop to weights ((20, 16, 32, 32), (1, 16, 28, 28), (1, 1), (1, 1), (1, 1)), # layer 1 backprop to weights
, ((60, 20, 28, 28), (10, 20, 5, 5), (1, 1), (2, 2), (1, 1)) # added a test case that fail from test_nnet.py.test_conv_nnet2 ((60, 20, 28, 28), (10, 20, 5, 5), (1, 1), (2, 2), (1, 1)), # added a test case that fail from test_nnet.py.test_conv_nnet2
, ((10, 5, 28, 28), (10, 5, 5, 5), (1, 1), (2, 2), (1, 1)) # test precedent but reduced that triger the error ((10, 5, 28, 28), (10, 5, 5, 5), (1, 1), (2, 2), (1, 1)), # test precedent but reduced that triger the error
# Test more than maxThreadsDim0 # Test more than maxThreadsDim0
, ((2, 4, 13, 1050), (3, 4, 10, 11), (1, 1), (1, 1), (1, 1)) ((2, 4, 13, 1050), (3, 4, 10, 11), (1, 1), (1, 1), (1, 1)),
, ((2, 4, 1050, 13), (3, 4, 10, 11), (1, 1), (1, 1), (1, 1)) ((2, 4, 1050, 13), (3, 4, 10, 11), (1, 1), (1, 1), (1, 1))
] ]
shapes += [ ((60, 1, 28, 28), (20, 1, 5, 5), (1, 1), (1, 1), (1, 1)) # test_lenet_28 1 layers shapes += [((60, 1, 28, 28), (20, 1, 5, 5), (1, 1), (1, 1), (1, 1)), # test_lenet_28 1 layers
, ((60, 20, 12, 12), (30, 20, 5, 5), (1, 1), (1, 1), (1, 1)) # test_lenet_28 2 layers ((60, 20, 12, 12), (30, 20, 5, 5), (1, 1), (1, 1), (1, 1)), # test_lenet_28 2 layers
, ((60, 30, 8, 8), (20, 30, 5, 5), (1, 1), (1, 1), (1, 1)) # test_lenet_28 bprop 1 full ((60, 30, 8, 8), (20, 30, 5, 5), (1, 1), (1, 1), (1, 1)), # test_lenet_28 bprop 1 full
, ((20, 60, 12, 12), (30, 60, 8, 8), (1, 1), (1, 1), (1, 1)) # test_lenet_28 bprop 2 valid ((20, 60, 12, 12), (30, 60, 8, 8), (1, 1), (1, 1), (1, 1)), # test_lenet_28 bprop 2 valid
# , ((1,60,28,28),(20,60,24,24), (1, 1), (1, 1), (1, 1))#test_lenet_28 bprop 2 valid # ((1,60,28,28),(20,60,24,24), (1, 1), (1, 1), (1, 1)), # test_lenet_28 bprop 2 valid
, ((10, 1, 64, 64), (20, 1, 7, 7), (1, 1), (1, 1), (1, 1)) # test_lenet_64 1 layers ((10, 1, 64, 64), (20, 1, 7, 7), (1, 1), (1, 1), (1, 1)), # test_lenet_64 1 layers
, ((10, 20, 29, 29), (30, 20, 7, 7), (1, 1), (1, 1), (1, 1)) # test_lenet_64 2 layers ((10, 20, 29, 29), (30, 20, 7, 7), (1, 1), (1, 1), (1, 1)), # test_lenet_64 2 layers
, ((10, 30, 23, 23), (20, 30, 7, 7), (1, 1), (1, 1), (1, 1)) # test_lenet_64 full ((10, 30, 23, 23), (20, 30, 7, 7), (1, 1), (1, 1), (1, 1)) # test_lenet_64 full
# , ((20,10,29,29),(30,10,23,23), (1, 1), (1, 1), (1, 1))#test_lenet_64 bprop 1 # ((20,10,29,29),(30,10,23,23), (1, 1), (1, 1), (1, 1)), # test_lenet_64 bprop 1
# , ((1,10,64,64),(20,10,58,58), (1, 1), (1, 1), (1, 1))#test_lenet_64 bprop 2 # ((1,10,64,64),(20,10,58,58), (1, 1), (1, 1), (1, 1)) # test_lenet_64 bprop 2
] ]
return shapes return shapes
...@@ -466,47 +464,46 @@ def _test_full(cls, mode=None, version=[-1], extra_shapes=[], ...@@ -466,47 +464,46 @@ def _test_full(cls, mode=None, version=[-1], extra_shapes=[],
shapes += [ shapes += [
# other test # other test
((2, 1, 2, 2), (1, 1, 2, 2), (1, 1), (1, 1), (1, 1)) ((2, 1, 2, 2), (1, 1, 2, 2), (1, 1), (1, 1), (1, 1)),
, ((3, 2, 4, 4), (4, 2, 4, 4), (1, 1), (1, 1), (1, 1)) ((3, 2, 4, 4), (4, 2, 4, 4), (1, 1), (1, 1), (1, 1)),
, ((4, 1, 10, 10), (1, 1, 2, 2), (1, 1), (1, 1), (1, 1)) ((4, 1, 10, 10), (1, 1, 2, 2), (1, 1), (1, 1), (1, 1)),
, ((1, 1, 4, 4), (1, 1, 2, 3), (1, 1), (1, 1), (1, 1)) ((1, 1, 4, 4), (1, 1, 2, 3), (1, 1), (1, 1), (1, 1)),
, ((4, 1, 10, 10), (1, 1, 2, 3), (1, 1), (1, 1), (1, 1)) ((4, 1, 10, 10), (1, 1, 2, 3), (1, 1), (1, 1), (1, 1)),
, ((4, 1, 10, 10), (1, 1, 2, 10), (1, 1), (1, 1), (1, 1)) ((4, 1, 10, 10), (1, 1, 2, 10), (1, 1), (1, 1), (1, 1)),
, ((4, 1, 20, 10), (1, 1, 2, 10), (1, 1), (1, 1), (1, 1)) ((4, 1, 20, 10), (1, 1, 2, 10), (1, 1), (1, 1), (1, 1)),
, ((3, 2, 8, 8), (4, 2, 4, 4), (1, 1), (1, 1), (1, 1)) # stack, nkern, bsize ((3, 2, 8, 8), (4, 2, 4, 4), (1, 1), (1, 1), (1, 1)), # stack, nkern, bsize
, ((3, 2, 8, 6), (4, 2, 4, 4), (1, 1), (1, 1), (1, 1)) # stack, nkern, bsize, non-square image ((3, 2, 8, 6), (4, 2, 4, 4), (1, 1), (1, 1), (1, 1)), # stack, nkern, bsize, non-square image
, ((3, 2, 8, 6), (4, 2, 4, 3), (1, 1), (1, 1), (1, 1)) # stack, nkern, bsize, non-square image, non-square kern ((3, 2, 8, 6), (4, 2, 4, 3), (1, 1), (1, 1), (1, 1)), # stack, nkern, bsize, non-square image, non-square kern
, ((3, 2, 8, 6), (4, 2, 4, 6), (1, 1), (1, 1), (1, 1)) # stack, nkern, bsize ,non-square image, non-square kern, kernsize==imgsize on one dim ((3, 2, 8, 6), (4, 2, 4, 6), (1, 1), (1, 1), (1, 1)), # stack, nkern, bsize ,non-square image, non-square kern, kernsize==imgsize on one dim
, ((16, 5, 64, 64), (8, 5, 8, 8), (1, 1), (1, 1), (1, 1)) # a big one ((16, 5, 64, 64), (8, 5, 8, 8), (1, 1), (1, 1), (1, 1)), # a big one
, ((16, 1, 28, 28), (20, 1, 5, 5), (1, 1), (1, 1), (1, 1)) # MNIST LeNET layer 1 ((16, 1, 28, 28), (20, 1, 5, 5), (1, 1), (1, 1), (1, 1)), # MNIST LeNET layer 1
, ((20, 16, 32, 32), (1, 16, 28, 28), (1, 1), (1, 1), (1, 1)) # layer 1 backprop to weights ((20, 16, 32, 32), (1, 16, 28, 28), (1, 1), (1, 1), (1, 1)) # layer 1 backprop to weights
] ]
if test_bigger_kernels: if test_bigger_kernels:
# Shapes where the kernel is larger than the image in some dimension # Shapes where the kernel is larger than the image in some dimension
shapes += [ shapes += [
((3, 1, 1, 1), (2, 1, 5, 3), (1, 1), (1, 1), (1, 1)) ((3, 1, 1, 1), (2, 1, 5, 3), (1, 1), (1, 1), (1, 1)),
, ((3, 2, 1, 1), (4, 2, 1, 1), (1, 1), (1, 1), (1, 1)) ((3, 2, 1, 1), (4, 2, 1, 1), (1, 1), (1, 1), (1, 1)),
, ((3, 2, 4, 4), (4, 2, 2, 6), (1, 1), (1, 1), (1, 1)) ((3, 2, 4, 4), (4, 2, 2, 6), (1, 1), (1, 1), (1, 1)),
, ((3, 2, 4, 4), (4, 2, 8, 6), (1, 1), (1, 1), (1, 1)) ((3, 2, 4, 4), (4, 2, 8, 6), (1, 1), (1, 1), (1, 1)),
, ((4, 2, 10, 10), (3, 2, 2, 12), (1, 1), (1, 1), (1, 1)) ((4, 2, 10, 10), (3, 2, 2, 12), (1, 1), (1, 1), (1, 1))
] ]
shapes += [ shapes += [((60, 1, 28, 28), (20, 1, 5, 5), (1, 1), (1, 1), (1, 1)), # test_lenet_28 1 layers
# ((60,1,28,28),(20,1,5,5), (1, 1), (1, 1), (1, 1))#test_lenet_28 1 layers # ((60, 20, 12, 12),(30, 20, 5, 5), (1, 1), (1, 1), (1, 1)), # test_lenet_28 2 layers
# , ((60,20,12,12),(30,20,5,5), (1, 1), (1, 1), (1, 1))#test_lenet_28 2 layers ((60, 30, 8, 8), (20, 30, 5, 5), (1, 1), (1, 1), (1, 1)), # test_lenet_28 bprop 1 full
((60, 30, 8, 8), (20, 30, 5, 5), (1, 1), (1, 1), (1, 1)) # test_lenet_28 bprop 1 full # ((20,60,12,12),(30,60,8,8), (1, 1), (1, 1), (1, 1)), # test_lenet_28 bprop 2 valid
# , ((20,60,12,12),(30,60,8,8), (1, 1), (1, 1), (1, 1))#test_lenet_28 bprop 2 valid # ((1,60,28,28),(20,60,24,24), (1, 1), (1, 1), (1, 1)), # test_lenet_28 bprop 2 valid
# , ((1,60,28,28),(20,60,24,24), (1, 1), (1, 1), (1, 1))#test_lenet_28 bprop 2 valid # ((10,1,64,64),(20,1,7,7), (1, 1), (1, 1), (1, 1)), # test_lenet_64 1 layers
# , ((10,1,64,64),(20,1,7,7), (1, 1), (1, 1), (1, 1))#test_lenet_64 1 layers # ((10,20,29,29),(30,20,7,7), (1, 1), (1, 1), (1, 1)), # test_lenet_64 2 layers
# , ((10,20,29,29),(30,20,7,7), (1, 1), (1, 1), (1, 1))#test_lenet_64 2 layers ((10, 30, 23, 23), (20, 30, 7, 7), (1, 1), (1, 1), (1, 1)), # test_lenet_64 full
, ((10, 30, 23, 23), (20, 30, 7, 7), (1, 1), (1, 1), (1, 1)) # test_lenet_64 full # ((20,10,29,29),(30,10,23,23), (1, 1), (1, 1), (1, 1)), # test_lenet_64 bprop 1
# , ((20,10,29,29),(30,10,23,23), (1, 1), (1, 1), (1, 1))#test_lenet_64 bprop 1 # ((1,10,64,64),(20,10,58,58), (1, 1), (1, 1), (1, 1)), # test_lenet_64 bprop 2
# , ((1,10,64,64),(20,10,58,58), (1, 1), (1, 1), (1, 1))#test_lenet_64 bprop 2
# Test more than maxThreadsDim0 # Test more than maxThreadsDim0
, ((2, 4, 13, 1050), (3, 4, 10, 11), (1, 1), (1, 1), (1, 1)) ((2, 4, 13, 1050), (3, 4, 10, 11), (1, 1), (1, 1), (1, 1)),
, ((2, 4, 1050, 13), (3, 4, 10, 11), (1, 1), (1, 1), (1, 1)) ((2, 4, 1050, 13), (3, 4, 10, 11), (1, 1), (1, 1), (1, 1)),
, ((1, 1, 44800, 1), (6, 1, 1, 1), (1, 1), (1, 1), (1, 1)) # This caused crash ((1, 1, 44800, 1), (6, 1, 1, 1), (1, 1), (1, 1), (1, 1)) # This caused crash
] ]
verbose = 0 verbose = 0
...@@ -636,7 +633,6 @@ class TestConv2DGPU(unittest.TestCase): ...@@ -636,7 +633,6 @@ class TestConv2DGPU(unittest.TestCase):
imshp_logical=featshp_logical[1:], imshp_logical=featshp_logical[1:],
kshp_logical=kshp[2:]) kshp_logical=kshp[2:])
def test_invalid_input_shape(self): def test_invalid_input_shape(self):
""" """
Tests that when the shape gived at build time is not the same as Tests that when the shape gived at build time is not the same as
...@@ -659,7 +655,7 @@ class TestConv2DGPU(unittest.TestCase): ...@@ -659,7 +655,7 @@ class TestConv2DGPU(unittest.TestCase):
for mode in ['valid', 'full']: for mode in ['valid', 'full']:
for shapes in [((3, 2, 8, 8), (4, 2, 5, 5), (8, 8)), for shapes in [((3, 2, 8, 8), (4, 2, 5, 5), (8, 8)),
((3, 2, 8, 8), (4, 2, 5, 5), (5, 8)), ((3, 2, 8, 8), (4, 2, 5, 5), (5, 8)),
#((3, 2, 8, 8), (4, 2, 5, 5), (8, 5)), # ((3, 2, 8, 8), (4, 2, 5, 5), (8, 5)),
# We use only the number of columns. # We use only the number of columns.
]: ]:
...@@ -777,8 +773,8 @@ def gemm_directly(bs, ch, nf, rImg1, rImg2, rFlt1, rFlt2, subsx, subsy, ...@@ -777,8 +773,8 @@ def gemm_directly(bs, ch, nf, rImg1, rImg2, rFlt1, rFlt2, subsx, subsy,
f = theano.function([i, k], op, mode=theano_mode) f = theano.function([i, k], op, mode=theano_mode)
gpuval = numpy.array(f( gpuval = numpy.array(f(
npy_img.transpose(1, 0, 2, 3), npy_img.transpose(1, 0, 2, 3),
npy_kern.transpose(1, 0, 2, 3)[:, :, ::-1, ::-1])).transpose( npy_kern.transpose(1, 0, 2, 3)[:, :, ::-1, ::-1])
1, 0, 2, 3) ).transpose(1, 0, 2, 3)
assert_allclose(cpuval, gpuval, rtol=1e-4) assert_allclose(cpuval, gpuval, rtol=1e-4)
...@@ -892,43 +888,43 @@ def benchmark(): ...@@ -892,43 +888,43 @@ def benchmark():
shapes_valid = [ shapes_valid = [
# test_lenet_28 shape # test_lenet_28 shape
((20, 60, 12, 12), (30, 60, 8, 8), (1, 1), (1, 1), (1, 1)) # valid ((20, 60, 12, 12), (30, 60, 8, 8), (1, 1), (1, 1), (1, 1)), # valid
, ((60, 20, 12, 12), (30, 20, 5, 5), (1, 1), (1, 1), (1, 1)) # valid ((60, 20, 12, 12), (30, 20, 5, 5), (1, 1), (1, 1), (1, 1)), # valid
, ((60, 1, 28, 28), (20, 1, 5, 5), (1, 1), (1, 1), (1, 1)) # valid ((60, 1, 28, 28), (20, 1, 5, 5), (1, 1), (1, 1), (1, 1)), # valid
, ((1, 60, 28, 28), (20, 60, 24, 24), (1, 1), (1, 1), (1, 1)) # valid ((1, 60, 28, 28), (20, 60, 24, 24), (1, 1), (1, 1), (1, 1)), # valid
# test_lenet_32 shape # test_lenet_32 shape
, ((20, 60, 14, 14), (30, 60, 10, 10), (1, 1), (1, 1), (1, 1)) # valid ((20, 60, 14, 14), (30, 60, 10, 10), (1, 1), (1, 1), (1, 1)), # valid
, ((60, 20, 14, 14), (30, 20, 5, 5), (1, 1), (1, 1), (1, 1)) # valid ((60, 20, 14, 14), (30, 20, 5, 5), (1, 1), (1, 1), (1, 1)), # valid
, ((60, 1, 32, 32), (20, 1, 5, 5), (1, 1), (1, 1), (1, 1)) # valid ((60, 1, 32, 32), (20, 1, 5, 5), (1, 1), (1, 1), (1, 1)), # valid
, ((1, 60, 32, 32), (20, 60, 28, 28), (1, 1), (1, 1), (1, 1)) # valid ((1, 60, 32, 32), (20, 60, 28, 28), (1, 1), (1, 1), (1, 1)), # valid
# test_lenet_64 shape # test_lenet_64 shape
, ((10, 20, 29, 29), (30, 20, 7, 7), (1, 1), (1, 1), (1, 1)) # valid ((10, 20, 29, 29), (30, 20, 7, 7), (1, 1), (1, 1), (1, 1)), # valid
, ((20, 10, 29, 29), (30, 10, 23, 23), (1, 1), (1, 1), (1, 1)) # valid ((20, 10, 29, 29), (30, 10, 23, 23), (1, 1), (1, 1), (1, 1)), # valid
, ((10, 1, 64, 64), (20, 1, 7, 7), (1, 1), (1, 1), (1, 1)) # valid ((10, 1, 64, 64), (20, 1, 7, 7), (1, 1), (1, 1), (1, 1)), # valid
, ((1, 10, 64, 64), (20, 10, 58, 58), (1, 1), (1, 1), (1, 1)) # valid ((1, 10, 64, 64), (20, 10, 58, 58), (1, 1), (1, 1), (1, 1)), # valid
# test_lenet_108 shape # test_lenet_108 shape
, ((10, 20, 51, 51), (30, 20, 7, 7), (1, 1), (1, 1), (1, 1)) # valid ((10, 20, 51, 51), (30, 20, 7, 7), (1, 1), (1, 1), (1, 1)), # valid
, ((20, 10, 51, 51), (30, 10, 45, 45), (1, 1), (1, 1), (1, 1)) # valid ((20, 10, 51, 51), (30, 10, 45, 45), (1, 1), (1, 1), (1, 1)), # valid
, ((10, 1, 108, 108), (20, 1, 7, 7), (1, 1), (1, 1), (1, 1)) # valid ((10, 1, 108, 108), (20, 1, 7, 7), (1, 1), (1, 1), (1, 1)), # valid
, ((1, 10, 108, 108), (20, 10, 102, 102), (1, 1), (1, 1), (1, 1)) # valid ((1, 10, 108, 108), (20, 10, 102, 102), (1, 1), (1, 1), (1, 1)), # valid
# test_lenet_256 shape # test_lenet_256 shape
, ((2, 20, 124, 124), (30, 20, 9, 9), (1, 1), (1, 1), (1, 1)) # valid ((2, 20, 124, 124), (30, 20, 9, 9), (1, 1), (1, 1), (1, 1)), # valid
, ((20, 2, 124, 124), (30, 2, 116, 116), (1, 1), (1, 1), (1, 1)) # valid ((20, 2, 124, 124), (30, 2, 116, 116), (1, 1), (1, 1), (1, 1)), # valid
, ((2, 1, 256, 256), (20, 1, 9, 9), (1, 1), (1, 1), (1, 1)) # valid ((2, 1, 256, 256), (20, 1, 9, 9), (1, 1), (1, 1), (1, 1)), # valid
, ((1, 2, 256, 256), (20, 2, 248, 248), (1, 1), (1, 1), (1, 1)) # valid ((1, 2, 256, 256), (20, 2, 248, 248), (1, 1), (1, 1), (1, 1)) # valid
] ]
shapes_full = [ shapes_full = [
# test_lenet_28 shape # test_lenet_28 shape
((60, 30, 8, 8), (20, 30, 5, 5), (1, 1), (1, 1), (1, 1)) # full ((60, 30, 8, 8), (20, 30, 5, 5), (1, 1), (1, 1), (1, 1)), # full
# test_lenet_32 shape # test_lenet_32 shape
, ((60, 30, 10, 10), (20, 30, 5, 5), (1, 1), (1, 1), (1, 1)) # full conv_full_patch_stack_padded' N=1 ((60, 30, 10, 10), (20, 30, 5, 5), (1, 1), (1, 1), (1, 1)), # full conv_full_patch_stack_padded' N=1
# test_lenet_64 shape # test_lenet_64 shape
, ((10, 30, 23, 23), (20, 30, 7, 7), (1, 1), (1, 1), (1, 1)) # full conv_full_patch_stack_padded' N=3 ((10, 30, 23, 23), (20, 30, 7, 7), (1, 1), (1, 1), (1, 1)), # full conv_full_patch_stack_padded' N=3
# test_lenet_108 shape # test_lenet_108 shape
, ((10, 30, 45, 45), (20, 30, 7, 7), (1, 1), (1, 1), (1, 1)) # full 'conv_full_patch_stack_padded' N=9 ((10, 30, 45, 45), (20, 30, 7, 7), (1, 1), (1, 1), (1, 1)), # full 'conv_full_patch_stack_padded' N=9
# test_lenet_256 shape # test_lenet_256 shape
, ((2, 30, 116, 116), (20, 30, 9, 9), (1, 1), (1, 1), (1, 1)) # full conv_reference_full ((2, 30, 116, 116), (20, 30, 9, 9), (1, 1), (1, 1), (1, 1)) # full conv_reference_full
] ]
version = [-1] version = [-1]
......
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