mirrors
/
PaddleClas
mirror of https://github.com/PaddlePaddle/PaddleClas.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

fix alexnet

pull/275/head
littletomatodonkey 2020-09-13 08:25:22 +00:00
parent 7ed80c5232
commit 3962b3853a
2 changed files with 91 additions and 61 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
ppcls/modeling/architectures/__init__.py
View File

@ -31,5 +31,6 @@ from .mobilenet_v1 import MobileNetV1_x0_25, MobileNetV1_x0_5, MobileNetV1_x0_75
from .mobilenet_v2 import MobileNetV2_x0_25, MobileNetV2_x0_5, MobileNetV2_x0_75, MobileNetV2, MobileNetV2_x1_5, MobileNetV2_x2_0 from .mobilenet_v2 import MobileNetV2_x0_25, MobileNetV2_x0_5, MobileNetV2_x0_75, MobileNetV2, MobileNetV2_x1_5, MobileNetV2_x2_0
from .mobilenet_v3 import MobileNetV3_small_x0_35, MobileNetV3_small_x0_5, MobileNetV3_small_x0_75, MobileNetV3_small_x1_0, MobileNetV3_small_x1_25, MobileNetV3_large_x0_35, MobileNetV3_large_x0_5, MobileNetV3_large_x0_75, MobileNetV3_large_x1_0, MobileNetV3_large_x1_25 from .mobilenet_v3 import MobileNetV3_small_x0_35, MobileNetV3_small_x0_5, MobileNetV3_small_x0_75, MobileNetV3_small_x1_0, MobileNetV3_small_x1_25, MobileNetV3_large_x0_35, MobileNetV3_large_x0_5, MobileNetV3_large_x0_75, MobileNetV3_large_x1_0, MobileNetV3_large_x1_25
from .shufflenet_v2 import ShuffleNetV2_x0_25, ShuffleNetV2_x0_33, ShuffleNetV2_x0_5, ShuffleNetV2, ShuffleNetV2_x1_5, ShuffleNetV2_x2_0, ShuffleNetV2_swish from .shufflenet_v2 import ShuffleNetV2_x0_25, ShuffleNetV2_x0_33, ShuffleNetV2_x0_5, ShuffleNetV2, ShuffleNetV2_x1_5, ShuffleNetV2_x2_0, ShuffleNetV2_swish
from .alexnet import AlexNet
from .distillation_models import ResNet50_vd_distill_MobileNetV3_large_x1_0 from .distillation_models import ResNet50_vd_distill_MobileNetV3_large_x1_0

145
ppcls/modeling/architectures/alexnet.py
View File

@ -1,103 +1,132 @@
import paddle import paddle
import paddle.fluid as fluid from paddle import ParamAttr
from paddle.fluid.param_attr import ParamAttr import paddle.nn as nn
from paddle.fluid.dygraph.nn import Conv2D, Pool2D, BatchNorm, Linear, Dropout import paddle.nn.functional as F
from paddle.nn import Conv2d, Pool2D, BatchNorm, Linear, Dropout, ReLU
from paddle.nn.initializer import Uniform
import math import math
__all__ = ["AlexNet"] __all__ = ["AlexNet"]
class ConvPoolLayer(fluid.dygraph.Layer):
class ConvPoolLayer(nn.Layer):
def __init__(self, def __init__(self,
inputc_channels, inputc_channels,
output_channels, output_channels,
filter_size, filter_size,
stride, stride,
padding, padding,
stdv, stdv,
groups=1, groups=1,
act=None, act=None,
name=None): name=None):
super(ConvPoolLayer, self).__init__() super(ConvPoolLayer, self).__init__()
self._conv = Conv2D(num_channels=inputc_channels, self.relu = ReLU() if act == "relu" else None
num_filters=output_channels,
filter_size=filter_size, self._conv = Conv2d(
stride=stride, in_channels=inputc_channels,
padding=padding, out_channels=output_channels,
groups=groups, kernel_size=filter_size,
param_attr=ParamAttr(name=name + "_weights", stride=stride,
initializer=fluid.initializer.Uniform(-stdv, stdv)), padding=padding,
bias_attr=ParamAttr(name=name + "_offset", groups=groups,
initializer=fluid.initializer.Uniform(-stdv, stdv)), weight_attr=ParamAttr(
act=act) name=name + "_weights", initializer=Uniform(-stdv, stdv)),
self._pool = Pool2D(pool_size=3, bias_attr=ParamAttr(
pool_stride=2, name=name + "_offset", initializer=Uniform(-stdv, stdv)))
pool_padding=0, self._pool = Pool2D(
pool_type="max") pool_size=3, pool_stride=2, pool_padding=0, pool_type="max")
def forward(self, inputs): def forward(self, inputs):
x = self._conv(inputs) x = self._conv(inputs)
if self.relu is not None:
x = self.relu(x)
x = self._pool(x) x = self._pool(x)
return x return x
class AlexNetDY(fluid.dygraph.Layer): class AlexNetDY(nn.Layer):
def __init__(self, class_dim=1000): def __init__(self, class_dim=1000):
super(AlexNetDY, self).__init__() super(AlexNetDY, self).__init__()
stdv = 1.0/math.sqrt(3*11*11) stdv = 1.0 / math.sqrt(3 * 11 * 11)
self._conv1 = ConvPoolLayer( self._conv1 = ConvPoolLayer(
3, 64, 11, 4, 2, stdv, act="relu", name="conv1") 3, 64, 11, 4, 2, stdv, act="relu", name="conv1")
stdv = 1.0/math.sqrt(64*5*5) stdv = 1.0 / math.sqrt(64 * 5 * 5)
self._conv2 = ConvPoolLayer( self._conv2 = ConvPoolLayer(
64, 192, 5, 1, 2, stdv, act="relu", name="conv2") 64, 192, 5, 1, 2, stdv, act="relu", name="conv2")
stdv = 1.0/math.sqrt(192*3*3) stdv = 1.0 / math.sqrt(192 * 3 * 3)
self._conv3 = Conv2D(192, 384, 3, stride=1, padding=1, self._conv3 = Conv2d(
param_attr=ParamAttr(name="conv3_weights", initializer=fluid.initializer.Uniform(-stdv, stdv)), 192,
bias_attr=ParamAttr(name="conv3_offset", initializer=fluid.initializer.Uniform(-stdv, stdv)), 384,
act="relu") 3,
stdv = 1.0/math.sqrt(384*3*3) stride=1,
self._conv4 = Conv2D(384, 256, 3, stride=1, padding=1, padding=1,
param_attr=ParamAttr(name="conv4_weights", initializer=fluid.initializer.Uniform(-stdv, stdv)), weight_attr=ParamAttr(
bias_attr=ParamAttr(name="conv4_offset", initializer=fluid.initializer.Uniform(-stdv, stdv)), name="conv3_weights", initializer=Uniform(-stdv, stdv)),
act="relu") bias_attr=ParamAttr(
stdv = 1.0/math.sqrt(256*3*3) name="conv3_offset", initializer=Uniform(-stdv, stdv)))
stdv = 1.0 / math.sqrt(384 * 3 * 3)
self._conv4 = Conv2d(
384,
256,
3,
stride=1,
padding=1,
weight_attr=ParamAttr(
name="conv4_weights", initializer=Uniform(-stdv, stdv)),
bias_attr=ParamAttr(
name="conv4_offset", initializer=Uniform(-stdv, stdv)))
stdv = 1.0 / math.sqrt(256 * 3 * 3)
self._conv5 = ConvPoolLayer( self._conv5 = ConvPoolLayer(
256, 256, 3, 1, 1, stdv, act="relu", name="conv5") 256, 256, 3, 1, 1, stdv, act="relu", name="conv5")
stdv = 1.0/math.sqrt(256*6*6) stdv = 1.0 / math.sqrt(256 * 6 * 6)
self._drop1 = Dropout(p=0.5) self._drop1 = Dropout(p=0.5)
self._fc6 = Linear(input_dim=256*6*6, self._fc6 = Linear(
output_dim=4096, in_features=256 * 6 * 6,
param_attr=ParamAttr(name="fc6_weights", initializer=fluid.initializer.Uniform(-stdv, stdv)), out_features=4096,
bias_attr=ParamAttr(name="fc6_offset", initializer=fluid.initializer.Uniform(-stdv, stdv)), weight_attr=ParamAttr(
act="relu") name="fc6_weights", initializer=Uniform(-stdv, stdv)),
bias_attr=ParamAttr(
name="fc6_offset", initializer=Uniform(-stdv, stdv)))
self._drop2 = Dropout(p=0.5) self._drop2 = Dropout(p=0.5)
self._fc7 = Linear(input_dim=4096, self._fc7 = Linear(
output_dim=4096, in_features=4096,
param_attr=ParamAttr(name="fc7_weights", initializer=fluid.initializer.Uniform(-stdv, stdv)), out_features=4096,
bias_attr=ParamAttr(name="fc7_offset", initializer=fluid.initializer.Uniform(-stdv, stdv)), weight_attr=ParamAttr(
act="relu") name="fc7_weights", initializer=Uniform(-stdv, stdv)),
self._fc8 = Linear(input_dim=4096, bias_attr=ParamAttr(
output_dim=class_dim, name="fc7_offset", initializer=Uniform(-stdv, stdv)))
param_attr=ParamAttr(name="fc8_weights", initializer=fluid.initializer.Uniform(-stdv, stdv)), self._fc8 = Linear(
bias_attr=ParamAttr(name="fc8_offset", initializer=fluid.initializer.Uniform(-stdv, stdv))) in_features=4096,
out_features=class_dim,
weight_attr=ParamAttr(
name="fc8_weights", initializer=Uniform(-stdv, stdv)),
bias_attr=ParamAttr(
name="fc8_offset", initializer=Uniform(-stdv, stdv)))
def forward(self, inputs): def forward(self, inputs):
x = self._conv1(inputs) x = self._conv1(inputs)
x = self._conv2(x) x = self._conv2(x)
x = self._conv3(x) x = self._conv3(x)
x = F.relu(x)
x = self._conv4(x) x = self._conv4(x)
x = F.relu(x)
x = self._conv5(x) x = self._conv5(x)
x = fluid.layers.flatten(x, axis=0) x = paddle.flatten(x, start_axis=1, stop_axis=-1)
x = self._drop1(x) x = self._drop1(x)
x = self._fc6(x) x = self._fc6(x)
x = F.relu(x)
x = self._drop2(x) x = self._drop2(x)
x = self._fc7(x) x = self._fc7(x)
x = F.relu(x)
x = self._fc8(x) x = self._fc8(x)
return x return x
def AlexNet(**args): def AlexNet(**args):
model = AlexNetDY(**args) model = AlexNetDY(**args)
return model return model