OSSForks/pytorch
1
0
Fork 0
mirror of https://github.com/zebrajr/pytorch.git synced 2025-12-07 12:21:27 +01:00
Code Issues Actions 38 Packages Projects Releases Wiki Activity

Use chainer-style constructor for Conv2d

Browse Source 
* Conv2d, MaxPool2d, and AvgPool2d have one argument for each of ksize,
   stride, and pad. This argument can be either a single number or a
   tuple of (h, w)
This commit is contained in:
Sam Gross 2016-09-07 15:51:44 -07:00
parent 1486d880b0
commit cd0929aa5e
6 changed files with 99 additions and 67 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

42
test/common_nn.py
View File

@ -9,7 +9,8 @@ try:
import torch.cuda
import torch.legacy.cunn
TEST_CUDA = True
except ImportError:
except Exception:
# TODO: catch ImportError once it works with "setup.py develop"
TEST_CUDA = False
PRECISION = 1e-5
@ -21,23 +22,6 @@ module_tests = [
input_size=(4, 10),
reference_fn=lambda i,p: torch.mm(i, p[0].t()) + p[1].view(1, -1).expand(4, 8)
),
dict(
module_name='Conv2d',
constructor_args=(3, 4, 3, 3),
input_size=(2, 3, 6, 6)
),
dict(
module_name='Conv2d',
constructor_args=(3, 4, 3, 3, 2, 2),
input_size=(2, 3, 6, 6),
desc='strided'
),
dict(
module_name='Conv2d',
constructor_args=(3, 4, 3, 3, 2, 2, 1, 1),
input_size=(2, 3, 6, 6),
desc='padding'
),
dict(
module_name='Threshold',
constructor_args=(2, 1),
@ -74,28 +58,6 @@ module_tests = [
module_name='Tanh',
input_size=(2, 3, 4, 5)
),
dict(
module_name='MaxPool2d',
constructor_args=(3, 3, 2, 2, 1, 1),
input_size=(1, 3, 7, 7)
),
dict(
module_name='AvgPool2d',
constructor_args=(2, 2),
input_size=(2, 3, 6, 6),
),
dict(
module_name='AvgPool2d',
constructor_args=(2, 2, 2, 2),
input_size=(2, 3, 6, 6),
desc='stride',
),
dict(
module_name='AvgPool2d',
constructor_args=(2, 2, 2, 2, 1, 1),
input_size=(2, 3, 6, 6),
desc='stride_pad',
),
dict(
module_name='Softmax',
input_size=(10, 20),

25
test/test_legacy_nn.py
View File

@ -343,6 +343,17 @@ tests = [
(1,),
input_size=[(1,), (2,), (3,), (4,)],
reference_fn=lambda i,_: i[1]),
OldModuleTest(nn.SpatialAveragePooling,
(2, 2),
input_size=(2, 3, 6, 6)),
OldModuleTest(nn.SpatialAveragePooling,
(2, 2, 2, 2),
input_size=(2, 3, 6, 6),
desc='stride'),
OldModuleTest(nn.SpatialAveragePooling,
(2, 2, 2, 2, 1, 1),
input_size=(2, 3, 6, 6),
desc='stride_pad'),
OldModuleTest(nn.SpatialAdaptiveMaxPooling,
(4, 4),
input_size=(2, 3, 8, 8)),
@ -352,6 +363,17 @@ tests = [
desc='irregular'),
# TODO: enable after implementing MaxPooling
# reference_fn=lambda i,_: nn.SpatialMaxPooling(2, 2).forward(i)),
OldModuleTest(nn.SpatialConvolution,
(3, 4, 3, 3),
input_size=(2, 3, 6, 6)),
OldModuleTest(nn.SpatialConvolution,
(3, 4, 3, 3, 2, 2),
input_size=(2, 3, 6, 6),
desc='strided'),
OldModuleTest(nn.SpatialConvolution,
(3, 4, 3, 3, 2, 2, 1, 1),
input_size=(2, 3, 6, 6),
desc='padding'),
OldModuleTest(nn.SpatialConvolutionLocal,
(3, 2, 4, 4, 2, 2),
input_size=(1, 3, 4, 4)),
@ -380,6 +402,9 @@ tests = [
(3, 2, 3, 3, 2, 2, 1, 1, 2, 2),
input_size=(2, 3, 8, 8),
desc='stride_pad'),
OldModuleTest(nn.SpatialMaxPooling,
(3, 3, 2, 2, 1, 1),
input_size=(1, 3, 7, 7)),
OldModuleTest(nn.SpatialReflectionPadding,
(1, 2, 3, 4),
input_size=(2, 3, 8, 8)),

46
test/test_nn.py
View File

@ -147,7 +147,7 @@ class TestNN(NNTestCase):
self.assertEqual(counter['backwards'], 7)
def test_volatile(self):
module = nn.Conv2d(2, 5, 3, 3, padh=1, padw=1)
module = nn.Conv2d(2, 5, ksize=3, pad=1)
input = torch.randn(1, 2, 10, 10)
x = Variable(input)
y = Variable(input.clone(), volatile=True)
@ -175,7 +175,49 @@ def add_test(test):
setattr(TestNN, cuda_test_name, lambda self,test=test: test.test_cuda(self))
for test_params in module_tests:
new_module_tests = [
dict(
module_name='Conv2d',
constructor_args=(3, 4, (3, 3)),
input_size=(2, 3, 6, 6)
),
dict(
module_name='Conv2d',
constructor_args=(3, 4, (3, 3), (2, 2)),
input_size=(2, 3, 6, 6),
desc='strided'
),
dict(
module_name='Conv2d',
constructor_args=(3, 4, (3, 3), (2, 2), (1, 1)),
input_size=(2, 3, 6, 6),
desc='padding'
),
dict(
module_name='MaxPool2d',
constructor_args=((3, 3), (2, 2), (1, 1)),
input_size=(1, 3, 7, 7)
),
dict(
module_name='AvgPool2d',
constructor_args=((2, 2),),
input_size=(2, 3, 6, 6),
),
dict(
module_name='AvgPool2d',
constructor_args=((2, 2), (2, 2)),
input_size=(2, 3, 6, 6),
desc='stride',
),
dict(
module_name='AvgPool2d',
constructor_args=((2, 2), (2, 2), (1, 1)),
input_size=(2, 3, 6, 6),
desc='stride_pad',
),
]
for test_params in module_tests + new_module_tests:
test_params = deepcopy(test_params)
# TODO: CUDA is not implemented yet
name = test_params.pop('module_name')

21
torch/nn/modules/conv.py
View File

@ -3,28 +3,31 @@ import torch
from torch.autograd import Variable
from .module import Module
from .utils import _pair
class Conv2d(Module):
def __init__(self, in_channels, out_channels, kh, kw, dh=1, dw=1, padh=0, padw=0):
def __init__(self, in_channels, out_channels, ksize, stride=1, pad=0,
nobias=False):
super(Conv2d, self).__init__()
self.in_channels = in_channels
self.out_channels = out_channels
self.kh = kh
self.kw = kw
self.dh = dh
self.dw = dw
self.padh = padh
self.padw = padw
self.kh, self.kw = _pair(ksize)
self.dh, self.dw = _pair(stride)
self.padh, self.padw = _pair(pad)
self.weight = Variable(torch.DoubleTensor(self.out_channels, self.in_channels, self.kh, self.kw))
self.bias = Variable(torch.DoubleTensor(self.out_channels))
if nobias:
self.bias = None
else:
self.bias = Variable(torch.DoubleTensor(self.out_channels))
self.reset_parameters()
def reset_parameters(self):
stdv = 1. / math.sqrt(self.kh * self.kw * self.in_channels)
self.weight.data.uniform_(-stdv, stdv)
self.bias.data.uniform_(-stdv, stdv)
if self.bias is not None:
self.bias.data.uniform_(-stdv, stdv)
def forward(self, input):
conv2d = self._backend.Conv2d(self.kw, self.kh, self.dw, self.dh, self.padw, self.padh)

26
torch/nn/modules/pooling.py
View File

@ -2,19 +2,16 @@ import torch
from torch.autograd import Variable
from .module import Module
from .utils import _pair
class MaxPool2d(Module):
def __init__(self, kh, kw, dh=None, dw=None, padh=0, padw=0, dilh=1, dilw=1, ceil_mode=False):
def __init__(self, ksize, stride=None, pad=0, dil=1, ceil_mode=False):
super(MaxPool2d, self).__init__()
self.kw = kw
self.kh = kh
self.dw = dw or kw
self.dh = dh or kh
self.padw = padw
self.padh = padh
self.dilh = dilh
self.dilw = dilw
self.kh, self.kw = _pair(ksize)
self.dh, self.dw = _pair(stride or ksize)
self.padh, self.padw = _pair(pad)
self.dilh, self.dilw = _pair(dil)
self.ceil_mode = ceil_mode
def forward(self, input):
@ -22,14 +19,11 @@ class MaxPool2d(Module):
class AvgPool2d(Module):
def __init__(self, kh, kw, dh=None, dw=None, padh=0, padw=0, ceil_mode=False, count_include_pad=True):
def __init__(self, ksize, stride=None, pad=0, ceil_mode=False, count_include_pad=True):
super(AvgPool2d, self).__init__()
self.kw = kw
self.kh = kh
self.dw = dw or kw
self.dh = dh or kh
self.padw = padw
self.padh = padh
self.kh, self.kw = _pair(ksize)
self.dh, self.dw = _pair(stride or ksize)
self.padh, self.padw = _pair(pad)
self.ceil_mode = ceil_mode
self.count_include_pad = count_include_pad

6
torch/nn/modules/utils.py Normal file
View File

@ -0,0 +1,6 @@
import collections
def _pair(x):
if isinstance(x, collections.Iterable):
return x
return x, x