Skip to content

P
PaddleClas

PaddlePaddle / PaddleClas
大约 1 年 前同步成功

通知 115
Star 4999
Fork 1114
P
PaddleClas
提交 f85ced36 编写于 作者: W wqz960
浏览文件
操作

fix

上级 0dfe15d2
隐藏空白更改
内联 并排
Showing with 68 addition and 78 deletion
ppcls/modeling/architectures/darknet.py
浏览文件 @ f85ced36
# coding=UTF-8
import numpy as np
import argparse
import paddle
......@@ -11,8 +10,6 @@ from paddle.fluid.dygraph.base import to_variable
from paddle.fluid import framework
import math
import sys
import time
__all__ = ["DarkNet53"]
......
ppcls/modeling/architectures/efficientnet.py
浏览文件 @ f85ced36
......@@ -11,8 +11,6 @@ from paddle.fluid.dygraph.base import to_variable
from paddle.fluid import framework
import math
import sys
import time
import collections
import re
import copy
......@@ -260,7 +258,7 @@ def _drop_connect(inputs, prob, is_test):
return output
class conv2d(fluid.dygraph.Layer):
class Conv2ds(fluid.dygraph.Layer):
def __init__(self,
input_channels,
output_channels,
......@@ -274,7 +272,7 @@ class conv2d(fluid.dygraph.Layer):
padding_type=None,
model_name=None,
cur_stage=None):
super(conv2d, self).__init__()
super(Conv2ds, self).__init__()
param_attr, bias_attr = initial_type(name=name, use_bias=use_bias)
......@@ -316,10 +314,6 @@ class conv2d(fluid.dygraph.Layer):
padding=padding,
param_attr=param_attr,
bias_attr=bias_attr)
# debug:
self.stride = stride
self.filter_size = filter_size
self.inps = inps
def forward(self, inputs):
x = self._conv(inputs)
......@@ -347,7 +341,7 @@ class ConvBNLayer(fluid.dygraph.Layer):
cur_stage=None):
super(ConvBNLayer, self).__init__()
self._conv = conv2d(
self._conv = Conv2ds(
input_channels=input_channels,
output_channels=output_channels,
filter_size=filter_size,
......@@ -383,7 +377,7 @@ class ConvBNLayer(fluid.dygraph.Layer):
return self._conv(inputs)
class Expand_Conv_Norm(fluid.dygraph.Layer):
class ExpandConvNorm(fluid.dygraph.Layer):
def __init__(self,
input_channels,
block_args,
......@@ -391,7 +385,7 @@ class Expand_Conv_Norm(fluid.dygraph.Layer):
name=None,
model_name=None,
cur_stage=None):
super(Expand_Conv_Norm, self).__init__()
super(ExpandConvNorm, self).__init__()
self.oup = block_args.input_filters * block_args.expand_ratio
self.expand_ratio = block_args.expand_ratio
......@@ -478,7 +472,7 @@ class Project_Conv_Norm(fluid.dygraph.Layer):
return self._conv(inputs)
class Se_Block(fluid.dygraph.Layer):
class SEBlock(fluid.dygraph.Layer):
def __init__(self,
input_channels,
num_squeezed_channels,
......@@ -487,11 +481,11 @@ class Se_Block(fluid.dygraph.Layer):
name=None,
model_name=None,
cur_stage=None):
super(Se_Block, self).__init__()
super(SEBlock, self).__init__()
self._pool = Pool2D(
pool_type="avg", global_pooling=True, use_cudnn=False)
self._conv1 = conv2d(
self._conv1 = Conv2ds(
input_channels,
num_squeezed_channels,
1,
......@@ -500,7 +494,7 @@ class Se_Block(fluid.dygraph.Layer):
act="swish",
name=name + "_se_reduce")
self._conv2 = conv2d(
self._conv2 = Conv2ds(
num_squeezed_channels,
oup,
1,
......@@ -517,7 +511,7 @@ class Se_Block(fluid.dygraph.Layer):
return fluid.layers.elementwise_mul(inputs, x)
class Mb_Conv_Block(fluid.dygraph.Layer):
class MbConvBlock(fluid.dygraph.Layer):
def __init__(self,
input_channels,
block_args,
......@@ -528,7 +522,7 @@ class Mb_Conv_Block(fluid.dygraph.Layer):
is_test=False,
model_name=None,
cur_stage=None):
super(Mb_Conv_Block, self).__init__()
super(MbConvBlock, self).__init__()
oup = block_args.input_filters * block_args.expand_ratio
self.block_args = block_args
......@@ -540,7 +534,7 @@ class Mb_Conv_Block(fluid.dygraph.Layer):
self.is_test = is_test
if self.expand_ratio != 1:
self._ecn = Expand_Conv_Norm(
self._ecn = ExpandConvNorm(
input_channels,
block_args,
padding_type=padding_type,
......@@ -559,7 +553,7 @@ class Mb_Conv_Block(fluid.dygraph.Layer):
if self.has_se:
num_squeezed_channels = max(
1, int(block_args.input_filters * block_args.se_ratio))
self._se = Se_Block(
self._se = SEBlock(
input_channels * block_args.expand_ratio,
num_squeezed_channels,
oup,
......@@ -587,14 +581,16 @@ class Mb_Conv_Block(fluid.dygraph.Layer):
if self.has_se:
x = self._se(x)
x = self._pcn(x)
if self.id_skip and self.block_args.stride == 1 and self.block_args.input_filters == self.block_args.output_filters:
if self.id_skip and \
self.block_args.stride == 1 and \
self.block_args.input_filters == self.block_args.output_filters:
if self.drop_connect_rate:
x = _drop_connect(x, self.drop_connect_rate, self.is_test)
x = fluid.layers.elementwise_add(x, inputs)
return x
class Conv_Stem_Norm(fluid.dygraph.Layer):
class ConvStemNorm(fluid.dygraph.Layer):
def __init__(self,
input_channels,
padding_type,
......@@ -602,7 +598,7 @@ class Conv_Stem_Norm(fluid.dygraph.Layer):
name=None,
model_name=None,
cur_stage=None):
super(Conv_Stem_Norm, self).__init__()
super(ConvStemNorm, self).__init__()
output_channels = round_filters(32, _global_params)
self._conv = ConvBNLayer(
......@@ -622,7 +618,7 @@ class Conv_Stem_Norm(fluid.dygraph.Layer):
return self._conv(inputs)
class Extract_Features(fluid.dygraph.Layer):
class ExtractFeatures(fluid.dygraph.Layer):
def __init__(self,
input_channels,
_block_args,
......@@ -631,11 +627,11 @@ class Extract_Features(fluid.dygraph.Layer):
use_se,
is_test,
model_name=None):
super(Extract_Features, self).__init__()
super(ExtractFeatures, self).__init__()
self._global_params = _global_params
self._conv_stem = Conv_Stem_Norm(
self._conv_stem = ConvStemNorm(
input_channels,
padding_type=padding_type,
_global_params=_global_params,
......@@ -673,7 +669,7 @@ class Extract_Features(fluid.dygraph.Layer):
_mc_block = self.add_sublayer(
"_blocks." + str(idx) + ".",
Mb_Conv_Block(
MbConvBlock(
block_args.input_filters,
block_args=block_args,
padding_type=padding_type,
......@@ -693,7 +689,7 @@ class Extract_Features(fluid.dygraph.Layer):
drop_connect_rate *= float(idx) / block_size
_mc_block = self.add_sublayer(
"block." + str(idx) + ".",
Mb_Conv_Block(
MbConvBlock(
block_args.input_filters,
block_args,
padding_type=padding_type,
......@@ -733,7 +729,7 @@ class EfficientNet(fluid.dygraph.Layer):
self.use_se = use_se
self.is_test = is_test
self._ef = Extract_Features(
self._ef = ExtractFeatures(
3,
self._block_args,
self._global_params,
......@@ -799,7 +795,7 @@ def EfficientNetB0_small(is_test=True,
use_se=False):
model = EfficientNet(
name='b0',
is_test=True,
is_test=is_test,
padding_type=padding_type,
override_params=override_params,
use_se=use_se)
......@@ -812,7 +808,7 @@ def EfficientNetB0(is_test=False,
use_se=True):
model = EfficientNet(
name='b0',
is_test=True,
is_test=is_test,
padding_type=padding_type,
override_params=override_params,
use_se=use_se)
......@@ -825,7 +821,7 @@ def EfficientNetB1(is_test=False,
use_se=True):
model = EfficientNet(
name='b1',
is_test=True,
is_test=is_test,
padding_type=padding_type,
override_params=override_params,
use_se=use_se)
......@@ -838,7 +834,7 @@ def EfficientNetB2(is_test=False,
use_se=True):
model = EfficientNet(
name='b2',
is_test=True,
is_test=is_test,
padding_type=padding_type,
override_params=override_params,
use_se=use_se)
......@@ -851,7 +847,7 @@ def EfficientNetB3(is_test=False,
use_se=True):
model = EfficientNet(
name='b3',
is_test=True,
is_test=is_test,
padding_type=padding_type,
override_params=override_params,
use_se=use_se)
......@@ -864,7 +860,7 @@ def EfficientNetB4(is_test=False,
use_se=True):
model = EfficientNet(
name='b4',
is_test=True,
is_test=is_test,
padding_type=padding_type,
override_params=override_params,
use_se=use_se)
......@@ -877,7 +873,7 @@ def EfficientNetB5(is_test=False,
use_se=True):
model = EfficientNet(
name='b5',
is_test=True,
is_test=is_test,
padding_type=padding_type,
override_params=override_params,
use_se=use_se)
......@@ -890,7 +886,7 @@ def EfficientNetB6(is_test=False,
use_se=True):
model = EfficientNet(
name='b6',
is_test=True,
is_test=is_test,
padding_type=padding_type,
override_params=override_params,
use_se=use_se)
......@@ -903,8 +899,10 @@ def EfficientNetB7(is_test=False,
use_se=True):
model = EfficientNet(
name='b7',
is_test=True,
is_test=is_test,
padding_type=padding_type,
override_params=override_params,
use_se=use_se)
return model
\ No newline at end of file
ppcls/modeling/architectures/googlenet.py
浏览文件 @ f85ced36
import numpy as np
import argparse
import paddle
import paddle.fluid as fluid
from paddle.fluid.param_attr import ParamAttr
......
ppcls/modeling/architectures/inception_v4.py
浏览文件 @ f85ced36
......@@ -8,10 +8,7 @@ from paddle.fluid.dygraph.nn import Conv2D, Pool2D, BatchNorm, Linear, Dropout
from paddle.fluid.dygraph.base import to_variable
from paddle.fluid import framework
import math
import sys
import time
__all__ = ["InceptionV4"]
......@@ -380,9 +377,9 @@ class InceptionC(fluid.dygraph.Layer):
return concat
class InceptionV4_DY(fluid.dygraph.Layer):
class InceptionV4DY(fluid.dygraph.Layer):
def __init__(self, class_dim=1000):
super(InceptionV4_DY, self).__init__()
super(InceptionV4DY, self).__init__()
self._inception_stem = Inception_Stem()
self._inceptionA_1 = InceptionA(name="1")
......@@ -445,5 +442,5 @@ class InceptionV4_DY(fluid.dygraph.Layer):
def InceptionV4():
model = InceptionV4_DY()
model = InceptionV4DY()
return model
ppcls/modeling/architectures/resnext101_wsl.py
浏览文件 @ f85ced36
......@@ -96,9 +96,9 @@ class Bottleneck_Block(fluid.dygraph.Layer):
y = self._short(inputs)
return fluid.layers.elementwise_add(x, y, act="relu")
class ResNeXt101_wsl(fluid.dygraph.Layer):
class ResNeXt101WSL(fluid.dygraph.Layer):
def __init__(self, layers=101, cardinality=32, width=48, class_dim=1000):
super(ResNeXt101_wsl, self).__init__()
super(ResNeXt101WSL, self).__init__()
self.class_dim = class_dim
......@@ -240,17 +240,17 @@ class ResNeXt101_wsl(fluid.dygraph.Layer):
return x
def ResNeXt101_32x8d_wsl():
model = ResNeXt101_wsl(cardinality=32, width=8)
model = ResNeXt101WSL(cardinality=32, width=8)
return model
def ResNeXt101_32x16d_wsl():
model = ResNeXt101_wsl(cardinality=32, width=16)
model = ResNeXt101WSL(cardinality=32, width=16)
return model
def ResNeXt101_32x32d_wsl():
model = ResNeXt101_wsl(cardinality=32, width=32)
model = ResNeXt101WSL(cardinality=32, width=32)
return model
def ResNeXt101_32x48d_wsl():
model = ResNeXt101_wsl(cardinality=32, width=48)
return model
model = ResNeXt101WSL(cardinality=32, width=48)
return model
\ No newline at end of file
ppcls/modeling/architectures/squeezenet.py
浏览文件 @ f85ced36
......@@ -15,14 +15,14 @@ import time
__all__ = ["SqueezeNet1_0", "SqueezeNet1_1"]
class Make_Fire_Conv(fluid.dygraph.Layer):
class MakeFireConv(fluid.dygraph.Layer):
def __init__(self,
input_channels,
output_channels,
filter_size,
padding=0,
name=None):
super(Make_Fire_Conv, self).__init__()
super(MakeFireConv, self).__init__()
self._conv = Conv2D(input_channels,
output_channels,
filter_size,
......@@ -34,23 +34,23 @@ class Make_Fire_Conv(fluid.dygraph.Layer):
def forward(self, inputs):
return self._conv(inputs)
class Make_Fire(fluid.dygraph.Layer):
class MakeFire(fluid.dygraph.Layer):
def __init__(self,
input_channels,
squeeze_channels,
expand1x1_channels,
expand3x3_channels,
name=None):
super(Make_Fire, self).__init__()
self._conv = Make_Fire_Conv(input_channels,
super(MakeFire, self).__init__()
self._conv = MakeFireConv(input_channels,
squeeze_channels,
1,
name=name + "_squeeze1x1")
self._conv_path1 = Make_Fire_Conv(squeeze_channels,
self._conv_path1 = MakeFireConv(squeeze_channels,
expand1x1_channels,
1,
name=name + "_expand1x1")
self._conv_path2 = Make_Fire_Conv(squeeze_channels,
self._conv_path2 = MakeFireConv(squeeze_channels,
expand3x3_channels,
3,
padding=1,
......@@ -78,16 +78,16 @@ class SqueezeNet(fluid.dygraph.Layer):
self._pool = Pool2D(pool_size=3,
pool_stride=2,
pool_type="max")
self._conv1 = Make_Fire(96, 16, 64, 64, name="fire2")
self._conv2 = Make_Fire(128, 16, 64, 64, name="fire3")
self._conv3 = Make_Fire(128, 32, 128, 128, name="fire4")
self._conv1 = MakeFire(96, 16, 64, 64, name="fire2")
self._conv2 = MakeFire(128, 16, 64, 64, name="fire3")
self._conv3 = MakeFire(128, 32, 128, 128, name="fire4")
self._conv4 = Make_Fire(256, 32, 128, 128, name="fire5")
self._conv5 = Make_Fire(256, 48, 192, 192, name="fire6")
self._conv6 = Make_Fire(384, 48, 192, 192, name="fire7")
self._conv7 = Make_Fire(384, 64, 256, 256, name="fire8")
self._conv4 = MakeFire(256, 32, 128, 128, name="fire5")
self._conv5 = MakeFire(256, 48, 192, 192, name="fire6")
self._conv6 = MakeFire(384, 48, 192, 192, name="fire7")
self._conv7 = MakeFire(384, 64, 256, 256, name="fire8")
self._conv8 = Make_Fire(512, 64, 256, 256, name="fire9")
self._conv8 = MakeFire(512, 64, 256, 256, name="fire9")
else:
self._conv = Conv2D(3,
64,
......@@ -100,16 +100,16 @@ class SqueezeNet(fluid.dygraph.Layer):
self._pool = Pool2D(pool_size=3,
pool_stride=2,
pool_type="max")
self._conv1 = Make_Fire(64, 16, 64, 64, name="fire2")
self._conv2 = Make_Fire(128, 16, 64, 64, name="fire3")
self._conv1 = MakeFire(64, 16, 64, 64, name="fire2")
self._conv2 = MakeFire(128, 16, 64, 64, name="fire3")
self._conv3 = Make_Fire(128, 32, 128, 128, name="fire4")
self._conv4 = Make_Fire(256, 32, 128, 128, name="fire5")
self._conv3 = MakeFire(128, 32, 128, 128, name="fire4")
self._conv4 = MakeFire(256, 32, 128, 128, name="fire5")
self._conv5 = Make_Fire(256, 48, 192, 192, name="fire6")
self._conv6 = Make_Fire(384, 48, 192, 192, name="fire7")
self._conv7 = Make_Fire(384, 64, 256, 256, name="fire8")
self._conv8 = Make_Fire(512, 64, 256, 256, name="fire9")
self._conv5 = MakeFire(256, 48, 192, 192, name="fire6")
self._conv6 = MakeFire(384, 48, 192, 192, name="fire7")
self._conv7 = MakeFire(384, 64, 256, 256, name="fire8")
self._conv8 = MakeFire(512, 64, 256, 256, name="fire9")
self._drop = Dropout(p=0.5)
self._conv9 = Conv2D(512,
......
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或