互联网站管理工作细则,创意小设计产品,深圳建设交易公司,软件开发项目计划神经网络-搭建小实战Sequential的使用 官网模型结构根据模型结构和数据的输入shape#xff0c;计算用在模型中的超参数coderunning log网络结构可视化 B站小土堆pytorch视频学习
官网
https://pytorch.org/docs/stable/generated/torch.nn.Sequential.html#torch.nn.Se… 神经网络-搭建小实战Sequential的使用 官网模型结构根据模型结构和数据的输入shape计算用在模型中的超参数coderunning log网络结构可视化 B站小土堆pytorch视频学习
官网
https://pytorch.org/docs/stable/generated/torch.nn.Sequential.html#torch.nn.Sequential
sequential 将模型结构组合起来 以逗号分割按顺序执行和compose使用方式类似。
模型结构 根据模型结构和数据的输入shape计算用在模型中的超参数 箭头指向部分还需要一层flatten层展开输入shape为一维
code
import torch
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Flatten, Linear, Sequential
from torch.utils.tensorboard import SummaryWriterclass MySeq(nn.Module):def __init__(self):super(MySeq, self).__init__()self.conv1 Conv2d(3, 32, kernel_size5, stride1, padding2)self.maxp1 MaxPool2d(2)self.conv2 Conv2d(32, 32, kernel_size5, stride1, padding2)self.maxp2 MaxPool2d(2)self.conv3 Conv2d(32, 64, kernel_size5, stride1, padding2)self.maxp3 MaxPool2d(2)self.flatten1 Flatten()self.linear1 Linear(1024, 64)self.linear2 Linear(64, 10)def forward(self, x):x self.conv1(x)x self.maxp1(x)x self.conv2(x)x self.maxp2(x)x self.conv3(x)x self.maxp3(x)x self.flatten1(x)x self.linear1(x)x self.linear2(x)return xclass MySeq2(nn.Module):def __init__(self):super(MySeq2, self).__init__()self.model1 Sequential(Conv2d(3, 32, kernel_size5, stride1, padding2),MaxPool2d(2),Conv2d(32, 32, kernel_size5, stride1, padding2),MaxPool2d(2),Conv2d(32, 64, kernel_size5, stride1, padding2),MaxPool2d(2),Flatten(),Linear(1024, 64),Linear(64, 10))def forward(self, x):x self.model1(x)return xmyseq MySeq()
input torch.ones(64, 3, 32, 32)
print(myseq)
print(input.shape)
output myseq(input)
print(output.shape)myseq2 MySeq2()
print(myseq2)
output2 myseq2(input)
print(output2.shape)wirter SummaryWriter(logs)
wirter.add_graph(myseq, input)
wirter.add_graph(myseq2, input)
running log
MySeq((conv1): Conv2d(3, 32, kernel_size(5, 5), stride(1, 1), padding(2, 2))(maxp1): MaxPool2d(kernel_size2, stride2, padding0, dilation1, ceil_modeFalse)(conv2): Conv2d(32, 32, kernel_size(5, 5), stride(1, 1), padding(2, 2))(maxp2): MaxPool2d(kernel_size2, stride2, padding0, dilation1, ceil_modeFalse)(conv3): Conv2d(32, 64, kernel_size(5, 5), stride(1, 1), padding(2, 2))(maxp3): MaxPool2d(kernel_size2, stride2, padding0, dilation1, ceil_modeFalse)(flatten1): Flatten(start_dim1, end_dim-1)(linear1): Linear(in_features1024, out_features64, biasTrue)(linear2): Linear(in_features64, out_features10, biasTrue)
)
torch.Size([64, 3, 32, 32])
torch.Size([64, 10])
MySeq2((model1): Sequential((0): Conv2d(3, 32, kernel_size(5, 5), stride(1, 1), padding(2, 2))(1): MaxPool2d(kernel_size2, stride2, padding0, dilation1, ceil_modeFalse)(2): Conv2d(32, 32, kernel_size(5, 5), stride(1, 1), padding(2, 2))(3): MaxPool2d(kernel_size2, stride2, padding0, dilation1, ceil_modeFalse)(4): Conv2d(32, 64, kernel_size(5, 5), stride(1, 1), padding(2, 2))(5): MaxPool2d(kernel_size2, stride2, padding0, dilation1, ceil_modeFalse)(6): Flatten(start_dim1, end_dim-1)(7): Linear(in_features1024, out_features64, biasTrue)(8): Linear(in_features64, out_features10, biasTrue))
)
torch.Size([64, 10])网络结构可视化
from torch.utils.tensorboard import SummaryWriter
wirter SummaryWriter(logs)
wirter.add_graph(myseq, input)tensorboard --logdirlogstensorboard 展示图文件, 双击每层网络可查看层定义细节
