小土堆-pytorch框架学习-P20-activate

非线性激活

padding层不进行介绍，原因：

1. 用不到
2. 卷积层有相关参数

---

batch_size是什么？

输入的最外层的一个维度。或者说第一个参数。

举例👇

import torch
input = torch.tensor([[1,-0.5],
					[-1,3]])
output = torch.reshape(input , (-1, 1,2,2))
#这里用-1是因为让机器自己算
#我们输入的多少自己知道，是一个，2×2的矩阵

以下是relu函数的使用instance👇

nn.relu.py
input = torch.tensor([[1,-0.5],
					  [-1,  3]
])
input = torch.reshape(input , (-1 , 1, 2,2))
print(f"input = {input}")

class Tudui(nn.Module):
	def __init__(self):
		super(Tudui,self).__init__()
		self.relu1 = nn.Relu(inplace=False)
    
    def forward(self,input):
        output = self.relu1()
        return output

tudui = Tudui()
output = tudui(input)
print(f"output = {output}")

这里的Relu函数中inplace是，这个意思：

inplace是以处理后的值是否取代原值。若设置为true，则取代原值；若设置为false，则不取代原值

in_1 = -1
out_1 = nn.ReLU(inplace = True)
#则
in_1 , out ==>1

in_2=-1
out_2 = nn.ReLU(inplace = False)
#则
in_2 ==> -1
out_2 ==> 0

---

import torch
import torchvision
import torch.nn
import tensorboardX.summary
from tensorboardX import SummaryWriter
from torch import nn
from torch.utils.data import DataLoader

input = torch.tensor([[1,-0.5],
                    [-1 ,3]])
input = torch.reshape(input , (-1, 1, 2, 2))

class Tudui(nn.Module):
    def __init__(self):
        super(Tudui, self).__init__()
        self.sigmoid1 = nn.Sigmoid()

    def forward(self,input):
        output  = self.sigmoid1(input)
        return output

tudui = Tudui()
output = tudui(input)
print(f" output = {output}")
print("="*111)

data_set = torchvision.datasets.CIFAR10(root="hymenoptera_data/val/CIFAR10",
                                        train = False,
                                        transform = torchvision.transforms.ToTensor())
data_loader = DataLoader(dataset = data_set , batch_size = 64)

writer = SummaryWriter("tb_logs")
step = 0
for data in data_loader:
    imgs , targets = data
    print(f"imgs = {imgs}")
    print("\n")
    writer.add_images("orignal_sigmoid" , imgs , global_step = step)

    output = tudui(imgs)
    print(f"output = {output}")
    print("\n")
    print("="*111)
    writer.add_images("final_sigmoid", output , global_step = step)

    step += 1
writer.close()

运行结果