目录
SummaryWriter 类 add_image() , add_scalar()使用
torchvision中数据集(DataLoader)的使用、与transforms相结合
二维卷积conv2d:F.conv2d(函数)和nn.conv2d(类)
优化器的使用(optim.zero_grad()、loss.backward()、optim.step())
-
dir()函数和help()的使用:
# dir()函数
dir(torch)
dir(torch.cuda)
dir(torch.cuda.is_available)
#help()函数 注意函数不不能加括号
help(torch.cuda.is_available)
-
Dataset类代码实践
PIL包 处理图片
from PIL import Image
img_path = "D:\\PycharmProjects\\learn_pytorch\\dataset\\train\\ants\\0013035.jpg"
img = Image.open(img_path)
img.size
# Out[6]: (768, 512)
img.show()
# 展示图片
os包 路径问题
dir_path = "dataset/train/ants"
import os
img_path_list = os.listdir(dir_path) # 将路径中的图片形成列表
# Return a list containing the names of the files in the directory.
img_path_list[0]
# Out[11]: '0013035.jpg'
读取数据
from torch.utils.data import Dataset
from PIL import Image
import os
class MyData(Dataset):
# 获取图片地址
def __init__(self, root_dir, label_dir):
self.root_dir = root_dir
self.label_dir = label_dir
self.path = os.path.join(self.root_dir , self.label_dir)
self.img_path = os.listdir(self.path)
#获取图片和label
def __getitem__(self, idx):
img_name = self.img_path[idx]
img_item_path = os.path.join(self.root_dir, self.label_dir, img_name)
img = Image.open(img_item_path)
label = self.label_dir
return img, label
# 返回数据集长度
def __len__(self):
return len(self.img_path)
root_dir = "dataset/train"
ants_label_dir = "ants"
bees_label_dir = "bees"
ants_dataset = MyData(root_dir, ants_label_dir)
bees_dataset = MyData(root_dir, bees_label_dir)
# 将数据集拼接
train_dataset = ants_dataset + bees_dataset
-
Tensorboard使用
SummaryWriter 类 add_image() , add_scalar()使用
from torch.utils.tensorboard import SummaryWriter
import numpy as np
from PIL import Image
writer = SummaryWriter("logs")
image_path = "data/train/ants_image/0013035.jpg"
img_PIL = Image.open(image_path)
img_array = np.array(img_PIL)
print(type(img_array))
print(img_array.shape)
writer.add_image("test", img_array, 2, dataformats='HWC') #官方说明:通道必须为HWC
# y = 2x ,可以每次删除一次log里面的文件
for i in range(0,100):
writer.add_scalar("y=2x",2 * i , i)
writer.close()
显示图片:
D:\PycharmProjects\learn_pytorch>tensorboard --logdir=logs
# Serving TensorBoard on localhost; to expose to the network, use a proxy or pass --bind_all
# TensorBoard 2.13.0 at http://localhost:6006/ (Press CTRL+C to quit)
## 可以改变上面端口6006
tensorboard --logdir=logs --port=6007 #改为6007
打开链接,展示:
Transforms 类
from PIL import Image
from torch.utils.tensorboard import SummaryWriter
from torchvision import transforms
# Python的用法 -》 tensor数据类型
img_path = "dataset/train/ants/6743948_2b8c096dda.jpg"
img_PIL = Image.open(img_path)
# print(img_PIL)
write = SummaryWriter("logs") # 会出现一个logs的文件夹目录
# 1. transforms如何使用
tensor_trans = transforms.ToTensor() # ToTensor是一个class , 在()中ctrl+p可以查看需要输入的类型
tensor_img = tensor_trans(img_PIL)
print(tensor_img)
write.add_image("tensor_img", tensor_img)
write.close()
运行结果:
常见的Trasnform:
常见transforms类的使用:
from PIL import Image
from torch.utils.tensorboard import SummaryWriter
from torchvision import transforms
writer = SummaryWriter("logs")
img = Image.open("image/pytorch.jpg") #PIL读取图片
print(img)
# ToTensor
trans_totensor = transforms.ToTensor()
img_tensor = trans_totensor(img)
writer.add_image("ToTensor", img_tensor)
# Normalize
print(img_tensor[0][0][0])
trans_norm = transforms.Normalize([0.4, 0.5, 0.8,0.2], [0.3, 0.5, 0.7,0.9]) #图片是RGBA 四维, A代表透明度
img_norm = trans_norm(img_tensor)
print(img_norm[0][0][0])
writer.add_image("trans_norm", img_norm, 2)
# Resize
print(img.size)
# img PIL -> resize() -> PIL
trans_resize = transforms.Resize((512, 512))
img_resize = trans_resize(img)
# img_resize PIL -> Totensor -> tensor类型
img_resize = trans_totensor(img_resize)
print(img_resize)
writer.add_image("trans_resize", img_resize, 0)
# Compose - resize - 2
trans_resize2 = transforms.Resize(512)
# PIL -> PIL -> tensor
trans_compose = transforms.Compose([trans_resize2, trans_totensor]) #compose方法
img_resize2 = trans_compose(img)
writer.add_image("trans_resize", img_resize2, 1)
# RandomCrop
trans_random = transforms.RandomCrop(50)
trans_compose2 = transforms.Compose([trans_random, trans_totensor])
for i in range(10):
img_crop = trans_compose2(img)
writer.add_image("RandomCrop", img_crop, i)
writer.close()
总结小记: 注意输入和输出类型;注意官方文档、所需参数;不确定返回值就用print(), print(type()), debug
torchvision中数据集(DataLoader)的使用、与transforms相结合
import torchvision
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
test_data = torchvision.datasets.CIFAR10("./dataset", train=False, transform=torchvision.transforms.ToTensor())
test_loader = DataLoader(dataset=test_data, batch_size=64, shuffle=True, num_workers=0, drop_last=False)
# 测试数据集中第一张图片及target
img, target = test_data[0]
print(img.shape)
print(target)
print(test_data.classes[target])
writer = SummaryWriter('dataloader')
for epoch in range(2):
step = 0
for data in test_loader:
imgs, target = data
# print(img.shape)
# print(target)
writer.add_images('epoch:{}'.format(epoch), imgs, step) #注意不是writer.add_image()
step = step + 1
# tensorboard显示完整step 可改为scalars
tensorboard --logdir=dataloader --samples_per_plugin=images=100000
神经网络(NN)
继承父类nn.Module
import torch
from torch import nn
class Tudui(nn.Module):
# code -> generate -> override methods
def __init__(self):
super().__init__() # 继承父类nn.Module的方法
def forward(self, input):
output = input + 1
return output
tudui = Tudui() #实例化
x1 = torch.tensor(1.0)
output1 = tudui(x1)
print(output1)
二维卷积conv2d:F.conv2d(函数)和nn.conv2d(类)
1、F.conv2d
import torch
import torch.nn.functional as F
input = torch.tensor([[1, 2, 0, 3, 1],
[0, 1, 2, 3, 1],
[1, 2, 1, 0, 0],
[5, 2, 3, 1, 1],
[2, 1, 0, 1, 1]])
kernel = torch.tensor([[1, 2, 1],
[0, 1, 0],
[2, 1, 0]])
input = torch.reshape(input, (1, 1, 5, 5))
kernel = torch.reshape(kernel, (1, 1, 3, 3))
print(input.shape)
print(kernel.shape)
output = F.conv2d(input, kernel, stride=1, padding=0)
print(output)
# print(output.shape) #torch.Size([1, 1, 3, 3])
2、nn.conv2d
import torch
import torchvision
from torch import nn
from torch.nn import Conv2d
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
dataset = torchvision.datasets.CIFAR10("./data", train=False, transform=torchvision.transforms.ToTensor(),
download=True)
dataloader = DataLoader(dataset, batch_size=64, shuffle=True, drop_last=True)
class Tudui(nn.Module): #们编写的模型所继承的nn.Module类中,其__call__方法内便包含了某种形式的对forward方法的调用,从而使得我们不需要显式地调用forward方法
def __init__(self):
super().__init__()
self.conv1 = Conv2d(in_channels=3, out_channels=6, kernel_size=3, stride=1, padding=0)
def forward(self, x):
x = self.conv1(x)
return x
tudui = Tudui()
# print(tudui)
writer = SummaryWriter("all_logs/P18")
step = 0
for data in dataloader:
imgs, targets = data
output = tudui(imgs)
print(output.shape)
writer.add_images("input", imgs, step)
output = torch.reshape(output, (-1, 3, 30, 30)) #通道为3才能显示彩色图像
writer.add_images("output", output, step)
step += 1
writer.close()
池化层(以MaxPool2d为例)
import torch
import torchvision
from torch import nn
from torch.nn import MaxPool2d
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
input = torch.tensor([[1, 2, 0, 3, 1],
[0, 1, 2, 3, 1],
[1, 2, 1, 0, 0],
[5, 2, 3, 1, 1],
[2, 1, 0, 1, 1]], dtype=float)
input = torch.reshape(input, (-1, 1, 5, 5))
# print(input.shape)
dataset = torchvision.datasets.CIFAR10("./data", train=False, transform=torchvision.transforms.ToTensor(),
download=True)
dataloader = DataLoader(dataset, batch_size=64, shuffle=False)
class maxpooling(nn.Module):
def __init__(self):
super().__init__()
self.maxpool1 = MaxPool2d(kernel_size=3, ceil_mode=True)
def forward(self, x):
x = self.maxpool1(x)
return x
mp = maxpooling()
# output = mp(input)
# print(mp)
# print(output)
writer = SummaryWriter("./all_logs/P19")
step = 0
for data in dataloader:
imgs, target = data
writer.add_images("P19_input", imgs, step)
imgs = mp(imgs)
# print(imgs)
writer.add_images("P19_output", imgs, step)
step += 1
writer.close()
非线性激活(ReLU、Sigmoid)
# 非线性激活
import torch
import torchvision.datasets
from torch import nn
from torch.nn import ReLU, Sigmoid
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
input = torch.randn((2, 2))
print(f'------input:------\n{input}')
input = torch.reshape(input, (-1, 1, 2, 2)) #-1自动计算batchsize,其余维度是1*2*2
print(input.shape)
dataset = torchvision.datasets.CIFAR10("./data", train=False, transform=torchvision.transforms.ToTensor(),
download=True)
dataloader = DataLoader(dataset, batch_size=64, shuffle=True)
# 直接写法
# relu = nn.ReLU() #注意nn.ReLU是一个类,与F.ReLU不同
# output = relu(input) #有无.forward都可以
#类写法
class fun(nn.Module):
def __init__(self):
super().__init__() #super继承nn.Module所有的东西,下一步可直接用ReLU
self.sigmoid1 = Sigmoid()
def forward(self, x):
x = self.sigmoid1(x)
return x
# relu1 = fun()
# output = relu1(input)
# print(f'\n\n-----output:------\n{output}')
sigmoid1 = fun()
writer = SummaryWriter("./all_logs/P20")
step = 0
for data in dataloader:
imgs, target = data
writer.add_images("P20relu_in", imgs, step)
output = sigmoid1(imgs)
writer.add_images("P20relu_out", output, step)
step += 1
writer.close()
线性层(linear layer)
import torch
import torchvision.datasets
from torch import nn
from torch.utils.data import DataLoader
dataset = torchvision.datasets.CIFAR10("data", train=False, transform=torchvision.transforms.ToTensor(),
download=True)
dataloader = DataLoader(dataset, batch_size=64, drop_last=True)
class linear(nn.Module):
def __init__(self):
super().__init__()
self.linear1 = nn.Linear(in_features=196608, out_features=10)
def forward(self, x):
y = self.linear1(x)
return y
linear1 = linear()
for data in dataloader:
imgs, target = data
print(imgs.shape)
# flatten = nn.Flatten()
# imgs = flatten(imgs) #torch.Size([64, 3072]) 这里的flatten考虑了batchsize
imgs = torch.flatten(imgs) #torch.Size([196608]) 没有batchsize,直接一维
print(imgs.shape)
imgs = linear1(imgs)
print(imgs.shape)
写网络模型(nn.Sequential用法)
import torch
import torch.nn as nn
class cifar10_model(nn.Module):
def __init__(self):
super().__init__()
self.model = nn.Sequential(
nn.Conv2d(3, 32, 5, stride=1, padding=2), # (n+2p-f)/s + 1 计算s和p
nn.MaxPool2d(2),
nn.Conv2d(32, 32, 5, stride=1, padding=2),
nn.MaxPool2d(2),
nn.Conv2d(32, 64, kernel_size=5, stride=1, padding=2),
nn.MaxPool2d(2),
nn.Flatten(), # input默认第0维是batchsize, 后三维形成全连接层
nn.Linear(64 * 4 * 4, 64),
nn.Linear(64, 10),
)
def forward(self, x):
x = self.model(x)
return x
model = cifar10_model()
print(model)
loss_function(network)
#loss function
#1、计算目标和输出之间的差距
#2、为更新输出提供依据(反向传播)grad
import torchvision.datasets
from torch import nn
from torch.utils.data import DataLoader
dataset = torchvision.datasets.CIFAR10("./data", train=False, transform=torchvision.transforms.ToTensor(),
download=True)
dataloader = DataLoader(dataset, batch_size=1)
class Cifar10_model(nn.Module):
def __init__(self):
super().__init__()
self.model = nn.Sequential(
nn.Conv2d(3, 32, 5, stride=1, padding=2), # (n+2p-f)/s + 1 计算s和p
nn.MaxPool2d(2),
nn.Conv2d(32, 32, 5, stride=1, padding=2),
nn.MaxPool2d(2),
nn.Conv2d(32, 64, kernel_size=5, stride=1, padding=2),
nn.MaxPool2d(2),
nn.Flatten(), # input默认第0维是batchsize, 后三维形成全连接层
nn.Linear(64 * 4 * 4, 64),
nn.Linear(64, 10),
)
def forward(self, x):
x = self.model(x)
return x
cifar10_model = Cifar10_model()
loss_cross = nn.CrossEntropyLoss()
for data in dataloader:
imgs, target = data
output = cifar10_model(imgs)
# print(imgs)
loss = loss_cross(output, target)
print(f'output:{output}\n'
f'target:{target}\n'
f'loss:{loss}\n')
loss.backward() #求梯度
print("ok")
优化器的使用(optim.zero_grad()、loss.backward()、optim.step())
import torch
import torchvision.datasets
from torch import nn
from torch.utils.data import DataLoader
dataset = torchvision.datasets.CIFAR10("./data", train=False, transform=torchvision.transforms.ToTensor(),
download=True)
dataloader = DataLoader(dataset, batch_size=1)
class Cifar10_model(nn.Module):
def __init__(self):
super().__init__()
self.model = nn.Sequential(
nn.Conv2d(3, 32, 5, stride=1, padding=2), # (n+2p-f)/s + 1 计算s和p
nn.MaxPool2d(2),
nn.Conv2d(32, 32, 5, stride=1, padding=2),
nn.MaxPool2d(2),
nn.Conv2d(32, 64, kernel_size=5, stride=1, padding=2),
nn.MaxPool2d(2),
nn.Flatten(), # input默认第0维是batchsize, 后三维形成全连接层
nn.Linear(64 * 4 * 4, 64),
nn.Linear(64, 10),
)
def forward(self, x):
x = self.model(x)
return x
cifar10_model = Cifar10_model()
loss_cross = nn.CrossEntropyLoss()
optim = torch.optim.SGD(params=cifar10_model.parameters(), lr=0.01)
for epoch in range(20):
running_loss = 0.0
for data in dataloader:
imgs, target = data
output = cifar10_model(imgs)
# print(imgs)
loss = loss_cross(output, target)
# print(f'output:{output}\n'
# f'target:{target}\n'
# f'loss:{loss}\n')
optim.zero_grad() #梯度值归零
loss.backward() #求梯度
optim.step() #参数更新
running_loss += loss
print(running_loss)
对现有网络进行修改(VGG16)
#对现有网络模型的修改 VGG16
import torchvision.datasets
from torch import nn
# train_data = torchvision.datasets.ImageNet("./data", split="train", download=True) 公开数据集无法下载
vgg16 = torchvision.models.vgg16()
vgg16_1 = torchvision.models.vgg16()
vgg16_2 = torchvision.models.vgg16()
vgg16_3 = torchvision.models.vgg16()
print(vgg16) # out_feature = 1000 , CIFAR10数据集输出为10
#在整个模型最后加
vgg16_1.add_module("add_linear", nn.Linear(1000, 10))
print(vgg16_1)
#在模型的classifier后加
vgg16_2.classifier.add_module("add_linear", nn.Linear(1000, 10))
print(vgg16_2)
#在原有模型中改
vgg16_3.classifier[6] = nn.Linear(4096, 10)
print(vgg16_3)
模型的保存与加载
# 模型保存
import torchvision
import torch
vgg16 = torchvision.models.vgg16()
# save方式一 :模型结构+模型参数
torch.save(vgg16, "./P26/vgg16_method1.pth")
# save方式二 :模型参数(官方推荐)
torch.save(vgg16.state_dict(), "./P26/vgg16_method2.pth")
# 模型加载
import torch
import torchvision.models
# load方式一
## 有陷阱 from P26_load import * (不用把模型重新写一遍)
model1 = torch.load("./P26/vgg16_method1.pth")
# print(model1)
# load方式二 (官方推荐)
model2 = torch.load("./P26/vgg16_method2.pth") # 只加载了参数
# print(model2)
vgg16 = torchvision.models.vgg16() # 把参数给到模型,一般这样用
vgg16.load_state_dict(torch.load("./P26/vgg16_method2.pth"))
print(vgg16)
模型的完整训练
利用cpu训练:
import torchvision.datasets
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
from P27_model import *
train_dataset = torchvision.datasets.CIFAR10("./data", train=True, transform=torchvision.transforms.ToTensor(),
download=True)
test_dataset = torchvision.datasets.CIFAR10("./data", train=False, transform=torchvision.transforms.ToTensor(),
download=False)
#数据长度 len
len_train = len(train_dataset)
len_test = len(test_dataset)
print(f"train_dataset长度:{len_train}")
print("test长度:{}".format(len_test))
# 数据加载器
train_dataloader = DataLoader(train_dataset, batch_size=64, shuffle=True, drop_last=False)
test_dataloader = DataLoader(test_dataset, batch_size=64, shuffle=True, drop_last=False)
# 创建模型
model = Cifar10_model()
# 损失函数
loss_fn = nn.CrossEntropyLoss()
# 优化器
learning_rate = 1e-2
optimizer = torch.optim.SGD(model.parameters(), lr=learning_rate)
# 设置训练网络的一些参数
total_train_step = 0
total_test_step = 0
epoch = 10
# 添加tensorboard
writer = SummaryWriter("./all_logs/P28_train")
for i in range(epoch):
print("------第{}轮训练开始:------".format(i+1))
# 训练开始
model.train() # 有特殊层的话要调用
for data in train_dataloader: # 50000张图片,一次data的batchsize为64, 单轮训练782次(看print)
imgs, target = data
output = model(imgs)
loss = loss_fn(output, target)
# 优化器优化模型
optimizer.zero_grad()
loss.backward()
optimizer.step()
total_train_step += 1
if total_train_step % 100 == 0:
print(f"训练次数:{total_train_step},loss:{loss.item()}") # .item() 例如可以将tensor数据类型转换为标量
writer.add_scalar("train_loss", loss, total_train_step)
# 测试模型
model.eval() # 有特殊层的话要调用
total_test_loss = 0
total_accuracy = 0
with torch.no_grad(): # 节约内存,提高速度
for data in test_dataloader:
imgs, target = data
output = model(imgs)
loss = loss_fn(output, target)
total_test_loss += loss.item()
accuracy = (output.argmax(1) == target).sum() # 1方向为横向
total_accuracy += accuracy
print("整体测试Loss:{}".format(total_test_loss))
print("整体测试集上的正确率:{}".format(total_accuracy / len_test))
writer.add_scalar("test_loss", total_test_loss, total_test_step)
writer.add_scalar("test_accuracy", total_accuracy / len_test, total_test_step)
total_test_step += 1
torch.save(model, "./P27_model/P27model_{}.pth".format(i)) # 第i轮次保存的模型
print("第{}轮模型已保存".format(i))
writer.close()
利用GPU训练
gpu训练方式一:
网络模型、损失函数、数据(输入、标注).cuda()
if torch.cuda.is_available():
xxxx = xxxx.cuda()
* gpu训练方式二:
设备: device = torch.device("cpu") or torch.device("cuda:0") 、 torch.device("cuda:1") 单显卡或者多显卡
# 简单写法
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
利用设备:xxx.to(device)
# 使用GPU训练模型
import torch.cuda
# 网络模型、、损失函数、数据(输入、标注) .to(device)
import torchvision.datasets
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
from P27_model import *
import time
train_dataset = torchvision.datasets.CIFAR10("./data", train=True, transform=torchvision.transforms.ToTensor(),
download=True)
test_dataset = torchvision.datasets.CIFAR10("./data", train=False, transform=torchvision.transforms.ToTensor(),
download=False)
#数据长度 len
len_train = len(train_dataset)
len_test = len(test_dataset)
print(f"train_dataset长度:{len_train}")
print("test长度:{}".format(len_test))
# 定义训练设备
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
# 数据加载器
train_dataloader = DataLoader(train_dataset, batch_size=64, shuffle=True, drop_last=False)
test_dataloader = DataLoader(test_dataset, batch_size=64, shuffle=True, drop_last=False)
# 创建模型
model = Cifar10_model()
model = model.to(device)
# 损失函数
loss_fn = nn.CrossEntropyLoss()
loss_fn = loss_fn.to(device)
# 优化器
learning_rate = 1e-2
optimizer = torch.optim.SGD(model.parameters(), lr=learning_rate)
# 设置训练网络的一些参数
total_train_step = 0
total_test_step = 0
epoch = 10
# 添加tensorboard
writer = SummaryWriter("./all_logs/P30_train_gpu1")
start_time = time.time()
for i in range(epoch):
print("------第{}轮训练开始:------".format(i+1))
# 训练开始
model.train() # 有特殊层的话要调用
for data in train_dataloader: # 50000张图片,一次data的batchsize为64, 单轮训练782次(看print)
imgs, target = data
imgs = imgs.to(device)
target = target.to(device)
output = model(imgs)
loss = loss_fn(output, target)
# 优化器优化模型
optimizer.zero_grad()
loss.backward()
optimizer.step()
total_train_step += 1
if total_train_step % 100 == 0:
print(f"训练次数:{total_train_step},loss:{loss.item()}") # .item() 例如可以将tensor数据类型转换为标量
end_time = time.time()
print(end_time - start_time)
writer.add_scalar("train_loss", loss, total_train_step)
# 测试模型
model.eval() # 有特殊层的话要调用
total_test_loss = 0
total_accuracy = 0
with torch.no_grad(): # 节约内存,提高速度
for data in test_dataloader:
imgs, target = data
imgs = imgs.to(device)
target = target.to(device)
output = model(imgs)
loss = loss_fn(output, target)
total_test_loss += loss.item()
accuracy = (output.argmax(1) == target).sum() # 1方向为横向
total_accuracy += accuracy
print("整体测试Loss:{}".format(total_test_loss))
print("整体测试集上的正确率:{}".format(total_accuracy / len_test))
writer.add_scalar("test_loss", total_test_loss, total_test_step)
writer.add_scalar("test_accuracy", total_accuracy / len_test, total_test_step)
total_test_step += 1
torch.save(model, "./P30_gpu1_train/P30_gpu1_model_{}.pth".format(i)) # 第i轮次保存的模型
print("第{}轮模型已保存".format(i+1))
writer.close()
完整的模型验证套路(测试、demo)
利用已经训练好的模型,给它提供输入。
# 完整模型验证套路(测试、demo)
import torch
import torchvision.transforms
from PIL import Image
# 有报错:Input type (torch.FloatTensor) and weight type (torch.cuda.FloatTensor) should be the same or input should be a MKLDNN tensor and weight is a dense tensor
# 使用gpu数据类型
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
img1_path = "image/dog1.png" # win中用 /
img1 = Image.open(img1_path)
# print(img1)
# img1 = img1.convert('RGB') # 如果是RGBA四通道(透明度),则需要转化为RGB
# print(img1)
transform = torchvision.transforms.Compose([torchvision.transforms.Resize((32, 32)),
torchvision.transforms.ToTensor()])
img1 = transform(img1).to(device)
# print(img1.shape)
model = torch.load("./P30_gpu1_train/P30_gpu1_model_9.pth")
print(model)
# RuntimeError: mat1 and mat2 shapes cannot be multiplied (64x16 and 1024x64)
# 错误:图片需要指定batchsize
img1 = torch.reshape(img1, (1, 3, 32, 32))
model.eval() # model测试
with torch.no_grad(): # 节约内存、性能
output = model(img1)
output = output.argmax(1)
print(output)
开源项目
运行开源项目时,train过程中的参数有required = True, 可以删掉并改为default = xxx
下面--dataroot --name --model 都是参数的名字,在源码中找到参数之后进行修改。