csdn_ggboy的博客

迁移学习迁移学习是在已有模型上进行学习

卷积神经网络学习手写数字# 定义超参数 input_size = 28 #图像的总尺寸28*28num_classes = 10 #标签的种类数num_epochs = 3 #训练的总循环周期batch_size = 64 #一个撮（批次）的大小，64张图片# 训练集train_dataset = datasets.MNIST(root='./data', train=True,

Mnist分类任务：网络基本构建与训练方法，常用函数解析torch.nn.functional模块nn.Module模块import torchx_train, y_train, x_valid, y_valid = map( torch.tensor, (x_train, y_train, x_valid, y_valid))#将数据转化为tensor类型import torch.nn.functional as Floss_func = F.cross_ent

tensor机器学习的基本单位x = torch.empty(3,5) x = torch.zeros(5,4)x = torch.rand(5,3) #0-1均匀分布的随机矩阵x = torch.randn(5,3)#正态分布（0,1）随机矩阵x = torch.randn_like(x)#和x大小一样的randn矩阵x = torch.tensor([[1,2],[3,4.3]])viewy = x.view(16)#将其变为16大小的tensorz = x.view(-1, 8)