微调¶
In [1]:
%matplotlib inline
import os
import torch
import torchvision
from torch import nn
from d2l import torch as d2l
热狗数据集来源于网络
In [3]:
d2l.DATA_HUB['hotdog'] = (d2l.DATA_URL + 'hotdog.zip',
'fba480ffa8aa7e0febbb511d181409f899b9baa5')
data_dir = d2l.download_extract('hotdog')
train_imgs = torchvision.datasets.ImageFolder(os.path.join(data_dir, 'train'))
test_imgs = torchvision.datasets.ImageFolder(os.path.join(data_dir, 'test'))
图像的大小和纵横比各有不同
In [4]:
hotdogs = [train_imgs[i][0] for i in range(8)]
not_hotdogs = [train_imgs[-i - 1][0] for i in range(8)]
d2l.show_images(hotdogs + not_hotdogs, 2, 8, scale=1.4);
数据增广
In [5]:
normalize = torchvision.transforms.Normalize([0.485, 0.456, 0.406],
[0.229, 0.224, 0.225])
train_augs = torchvision.transforms.Compose([
torchvision.transforms.RandomResizedCrop(224),
torchvision.transforms.RandomHorizontalFlip(),
torchvision.transforms.ToTensor(), normalize])
test_augs = torchvision.transforms.Compose([
torchvision.transforms.Resize(256),
torchvision.transforms.CenterCrop(224),
torchvision.transforms.ToTensor(), normalize])
定义和初始化模型
In [7]:
pretrained_net = torchvision.models.resnet18(pretrained=True)
pretrained_net.fc
Out[7]:
Linear(in_features=512, out_features=1000, bias=True)
In [8]:
finetune_net = torchvision.models.resnet18(pretrained=True)
finetune_net.fc = nn.Linear(finetune_net.fc.in_features, 2)
nn.init.xavier_uniform_(finetune_net.fc.weight);
微调模型
In [9]:
def train_fine_tuning(net, learning_rate, batch_size=128, num_epochs=5,
param_group=True):
train_iter = torch.utils.data.DataLoader(
torchvision.datasets.ImageFolder(os.path.join(data_dir, 'train'),
transform=train_augs),
batch_size=batch_size, shuffle=True)
test_iter = torch.utils.data.DataLoader(
torchvision.datasets.ImageFolder(os.path.join(data_dir, 'test'),
transform=test_augs),
batch_size=batch_size)
devices = d2l.try_all_gpus()
loss = nn.CrossEntropyLoss(reduction="none")
if param_group:
params_1x = [
param for name, param in net.named_parameters()
if name not in ["fc.weight", "fc.bias"]]
trainer = torch.optim.SGD([{
'params': params_1x}, {
'params': net.fc.parameters(),
'lr': learning_rate * 10}], lr=learning_rate,
weight_decay=0.001)
else:
trainer = torch.optim.SGD(net.parameters(), lr=learning_rate,
weight_decay=0.001)
d2l.train_ch13(net, train_iter, test_iter, loss, trainer, num_epochs,
devices)
使用较小的学习率
In [10]:
train_fine_tuning(finetune_net, 5e-5)
loss 0.263, train acc 0.905, test acc 0.934 841.8 examples/sec on [device(type='cuda', index=0), device(type='cuda', index=1)]
为了进行比较， 所有模型参数初始化为随机值
In [11]:
scratch_net = torchvision.models.resnet18()
scratch_net.fc = nn.Linear(scratch_net.fc.in_features, 2)
train_fine_tuning(scratch_net, 5e-4, param_group=False)
loss 0.416, train acc 0.819, test acc 0.750 1570.1 examples/sec on [device(type='cuda', index=0), device(type='cuda', index=1)]
In [13]:
for param in finetune_net.parameters():
param.requires_grad = False
weight = pretrained_net.fc.weight
hotdog_w = torch.split(weight.data, 1, dim=0)[713]
hotdog_w.shape
Out[13]:
torch.Size([1, 512])
微调
In [1]:
%matplotlib inline
import os
import torch
import torchvision
from torch import nn
from d2l import torch as d2l