YaResNet.
Yet Another ResNet model.
from model_constructor.yaresnet import YaBasicBlock, YaBottleneckBlock
YaBasicBlock, YaBottleneckBlock
bl = YaBasicBlock(64, 64)
bl
output
YaBasicBlock(
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): ReLU(inplace=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(merge): ReLU(inplace=True)
)
pool = partial(nn.AvgPool2d, kernel_size=2, ceil_mode=True)
bl = YaBottleneckBlock(64, 128, stride=2, pool=pool, act_fn=nn.LeakyReLU, bn_1st=False)
bl
output
YaBottleneckBlock(
(reduce): AvgPool2d(kernel_size=2, stride=2, padding=0)
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(64, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)
(act_fn): LeakyReLU(negative_slope=0.01, inplace=True)
(bn): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(act_fn): LeakyReLU(negative_slope=0.01, inplace=True)
(bn): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(conv_2): ConvBnAct(
(conv): Conv2d(32, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(id_conv): ConvBnAct(
(conv): Conv2d(64, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(merge): LeakyReLU(negative_slope=0.01, inplace=True)
)
bl = YaBottleneckBlock(
64,
128,
expansion=2,
stride=2,
pool=pool,
act_fn=nn.LeakyReLU,
bn_1st=False,
groups=4,
)
bl
output
YaBottleneckBlock(
(reduce): AvgPool2d(kernel_size=2, stride=2, padding=0)
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(act_fn): LeakyReLU(negative_slope=0.01, inplace=True)
(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=4, bias=False)
(act_fn): LeakyReLU(negative_slope=0.01, inplace=True)
(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(conv_2): ConvBnAct(
(conv): Conv2d(64, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(id_conv): ConvBnAct(
(conv): Conv2d(64, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(merge): LeakyReLU(negative_slope=0.01, inplace=True)
)
bl = YaBottleneckBlock(
64,
128,
stride=2,
pool=pool,
act_fn=nn.LeakyReLU,
bn_1st=False,
div_groups=4,
)
bl
output
YaBottleneckBlock(
(reduce): AvgPool2d(kernel_size=2, stride=2, padding=0)
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(64, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)
(act_fn): LeakyReLU(negative_slope=0.01, inplace=True)
(bn): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=8, bias=False)
(act_fn): LeakyReLU(negative_slope=0.01, inplace=True)
(bn): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(conv_2): ConvBnAct(
(conv): Conv2d(32, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(id_conv): ConvBnAct(
(conv): Conv2d(64, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(merge): LeakyReLU(negative_slope=0.01, inplace=True)
)
bl = YaBasicBlock(
64,
128,
stride=2,
pool=pool,
act_fn=nn.Mish,
bn_1st=False,
dw=True,
)
bl
output
YaBasicBlock(
(reduce): AvgPool2d(kernel_size=2, stride=2, padding=0)
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
(act_fn): Mish(inplace=True)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=128, bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(id_conv): ConvBnAct(
(conv): Conv2d(64, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(merge): Mish(inplace=True)
)
Se, Sa
from model_constructor.layers import SimpleSelfAttention, SEModule
bl = YaBottleneckBlock(
64,
128,
stride=2,
pool=pool,
act_fn=nn.GELU,
dw=True,
se=SEModule,
)
bl
output
YaBottleneckBlock(
(reduce): AvgPool2d(kernel_size=2, stride=2, padding=0)
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(64, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): GELU(approximate='none')
)
(conv_1): ConvBnAct(
(conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)
(bn): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): GELU(approximate='none')
)
(conv_2): ConvBnAct(
(conv): Conv2d(32, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(se): SEModule(
(squeeze): AdaptiveAvgPool2d(output_size=1)
(excitation): Sequential(
(reduce): Linear(in_features=128, out_features=8, bias=True)
(se_act): ReLU(inplace=True)
(expand): Linear(in_features=8, out_features=128, bias=True)
(se_gate): Sigmoid()
)
)
)
(id_conv): ConvBnAct(
(conv): Conv2d(64, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(merge): GELU(approximate='none')
)
bl = YaBottleneckBlock(
64,
128,
stride=2,
pool=pool,
act_fn=nn.LeakyReLU,
dw=True,
sa=SimpleSelfAttention,
)
bl
output
YaBottleneckBlock(
(reduce): AvgPool2d(kernel_size=2, stride=2, padding=0)
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(64, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): LeakyReLU(negative_slope=0.01, inplace=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)
(bn): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): LeakyReLU(negative_slope=0.01, inplace=True)
)
(conv_2): ConvBnAct(
(conv): Conv2d(32, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(sa): SimpleSelfAttention(
(conv): Conv1d(128, 128, kernel_size=(1,), stride=(1,), bias=False)
)
)
(id_conv): ConvBnAct(
(conv): Conv2d(64, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(merge): LeakyReLU(negative_slope=0.01, inplace=True)
)
bl = YaBottleneckBlock(
64,
128,
expansion=4,
stride=2,
pool=pool,
act_fn=nn.LeakyReLU,
dw=True,
se=SEModule,
sa=SimpleSelfAttention)
bl
output
YaBottleneckBlock(
(reduce): AvgPool2d(kernel_size=2, stride=2, padding=0)
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(64, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): LeakyReLU(negative_slope=0.01, inplace=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)
(bn): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): LeakyReLU(negative_slope=0.01, inplace=True)
)
(conv_2): ConvBnAct(
(conv): Conv2d(32, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(se): SEModule(
(squeeze): AdaptiveAvgPool2d(output_size=1)
(excitation): Sequential(
(reduce): Linear(in_features=128, out_features=8, bias=True)
(se_act): ReLU(inplace=True)
(expand): Linear(in_features=8, out_features=128, bias=True)
(se_gate): Sigmoid()
)
)
(sa): SimpleSelfAttention(
(conv): Conv1d(128, 128, kernel_size=(1,), stride=(1,), bias=False)
)
)
(id_conv): ConvBnAct(
(conv): Conv2d(64, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(merge): LeakyReLU(negative_slope=0.01, inplace=True)
)
YaResNet constructor.
from model_constructor import ModelConstructor
from model_constructor.xresnet import xresnet_stem
yaresnet = ModelConstructor(
block=YaBasicBlock,
make_stem=xresnet_stem,
stem_sizes=[3, 32, 64, 64],
name='YaResNet',
)
yaresnet
output
YaResNet
in_chans: 3, num_classes: 1000
expansion: 1, groups: 1, dw: False, div_groups: None
act_fn: ReLU, sa: False, se: False
stem sizes: [3, 32, 64, 64], stride on 0
body sizes [64, 128, 256, 512]
layers: [2, 2, 2, 2]
yaresnet.print_changed_fields()
output
Changed fields:
name: YaResNet
block: YaBasicBlock
stem_sizes: [3, 32, 64, 64]
make_stem: xresnet_stem
yaresnet.stem
output
Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(3, 3, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn): BatchNorm2d(3, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): ReLU(inplace=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(3, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): ReLU(inplace=True)
)
(conv_2): ConvBnAct(
(conv): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): ReLU(inplace=True)
)
(conv_3): ConvBnAct(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): ReLU(inplace=True)
)
(stem_pool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)
)
yaresnet.body
output
Sequential(
(l_0): Sequential(
(bl_0): YaBasicBlock(
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): ReLU(inplace=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(merge): ReLU(inplace=True)
)
(bl_1): YaBasicBlock(
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): ReLU(inplace=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(merge): ReLU(inplace=True)
)
)
(l_1): Sequential(
(bl_0): YaBasicBlock(
(reduce): ConvBnAct(
(conv): Conv2d(64, 64, kernel_size=(1, 1), stride=(2, 2), bias=False)
(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): ReLU(inplace=True)
)
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): ReLU(inplace=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(id_conv): ConvBnAct(
(conv): Conv2d(64, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(merge): ReLU(inplace=True)
)
(bl_1): YaBasicBlock(
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): ReLU(inplace=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(merge): ReLU(inplace=True)
)
)
(l_2): Sequential(
(bl_0): YaBasicBlock(
(reduce): ConvBnAct(
(conv): Conv2d(128, 128, kernel_size=(1, 1), stride=(2, 2), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): ReLU(inplace=True)
)
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): ReLU(inplace=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(id_conv): ConvBnAct(
(conv): Conv2d(128, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(merge): ReLU(inplace=True)
)
(bl_1): YaBasicBlock(
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): ReLU(inplace=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(merge): ReLU(inplace=True)
)
)
(l_3): Sequential(
(bl_0): YaBasicBlock(
(reduce): ConvBnAct(
(conv): Conv2d(256, 256, kernel_size=(1, 1), stride=(2, 2), bias=False)
(bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): ReLU(inplace=True)
)
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): ReLU(inplace=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(id_conv): ConvBnAct(
(conv): Conv2d(256, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(merge): ReLU(inplace=True)
)
(bl_1): YaBasicBlock(
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): ReLU(inplace=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(merge): ReLU(inplace=True)
)
)
)
yaresnet.head
output
Sequential(
(pool): AdaptiveAvgPool2d(output_size=1)
(flat): Flatten(start_dim=1, end_dim=-1)
(fc): Linear(in_features=512, out_features=1000, bias=True)
)
Lots of experiments showed that it worth trying Mish activation function.
yaresnet.act_fn = torch.nn.Mish
yaresnet()
output
YaResNet(
block: YaBasicBlock, act_fn: Mish, stem_sizes: [3, 32, 64, 64], make_stem: xresnet_stem
(stem): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(3, 3, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn): BatchNorm2d(3, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): Mish(inplace=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(3, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): Mish(inplace=True)
)
(conv_2): ConvBnAct(
(conv): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): Mish(inplace=True)
)
(conv_3): ConvBnAct(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): Mish(inplace=True)
)
(stem_pool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)
)
(body): Sequential(
(l_0): Sequential(
(bl_0): YaBasicBlock(
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): Mish(inplace=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(merge): Mish(inplace=True)
)
(bl_1): YaBasicBlock(
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): Mish(inplace=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(merge): Mish(inplace=True)
)
)
(l_1): Sequential(
(bl_0): YaBasicBlock(
(reduce): ConvBnAct(
(conv): Conv2d(64, 64, kernel_size=(1, 1), stride=(2, 2), bias=False)
(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): ReLU(inplace=True)
)
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): Mish(inplace=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(id_conv): ConvBnAct(
(conv): Conv2d(64, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(merge): Mish(inplace=True)
)
(bl_1): YaBasicBlock(
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): Mish(inplace=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(merge): Mish(inplace=True)
)
)
(l_2): Sequential(
(bl_0): YaBasicBlock(
(reduce): ConvBnAct(
(conv): Conv2d(128, 128, kernel_size=(1, 1), stride=(2, 2), bias=False)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): ReLU(inplace=True)
)
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): Mish(inplace=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(id_conv): ConvBnAct(
(conv): Conv2d(128, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(merge): Mish(inplace=True)
)
(bl_1): YaBasicBlock(
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): Mish(inplace=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(merge): Mish(inplace=True)
)
)
(l_3): Sequential(
(bl_0): YaBasicBlock(
(reduce): ConvBnAct(
(conv): Conv2d(256, 256, kernel_size=(1, 1), stride=(2, 2), bias=False)
(bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): ReLU(inplace=True)
)
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): Mish(inplace=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(id_conv): ConvBnAct(
(conv): Conv2d(256, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(merge): Mish(inplace=True)
)
(bl_1): YaBasicBlock(
(convs): Sequential(
(conv_0): ConvBnAct(
(conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act_fn): Mish(inplace=True)
)
(conv_1): ConvBnAct(
(conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(merge): Mish(inplace=True)
)
)
)
(head): Sequential(
(pool): AdaptiveAvgPool2d(output_size=1)
(flat): Flatten(start_dim=1, end_dim=-1)
(fc): Linear(in_features=512, out_features=1000, bias=True)
)
)
yaresnet.se = SEModule
yaresnet.body.l_0.bl_0
YaResnet34, YaResnet50
We has Resnet34 and Resnet50 like models predefined, we can impoer it as:
from model_constructor.yaresnet import YaResNet34, YaResNet50
But lets create it.
class YaResNet(ModelConstructor):
make_stem: Callable[[ModelCfg], ModSeq] = xresnet_stem
stem_sizes: list[int] = [32, 64, 64]
block: type[nn.Module] = YaBasicBlock
act_fn: type[nn.Module] = nn.Mish
pool: Optional[Callable[[Any], nn.Module]] = partial(
nn.AvgPool2d, kernel_size=2, ceil_mode=True
)
class YaResNet34(YaResNet):
layers: list[int] = [3, 4, 6, 3]
yaresnet34 = YaResNet34()
yaresnet34
output
YaResNet34
in_chans: 3, num_classes: 1000
expansion: 1, groups: 1, dw: False, div_groups: None
act_fn: Mish, sa: False, se: False
stem sizes: [32, 64, 64], stride on 0
body sizes [64, 128, 256, 512]
layers: [3, 4, 6, 3]
class YaResNet50(YaResNet34):
block: type[nn.Module] = YaBottleneckBlock
block_sizes: list[int] = [256, 512, 1024, 2048]
yaresnet50 = YaResNet50()
yaresnet50
output
YaResNet50
in_chans: 3, num_classes: 1000
expansion: 1, groups: 1, dw: False, div_groups: None
act_fn: Mish, sa: False, se: False
stem sizes: [32, 64, 64], stride on 0
body sizes [256, 512, 1024, 2048]
layers: [3, 4, 6, 3]
model_constructor by ayasyrev