Tensorboard SummaryWriter()

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132

import torch import torch.nn as nn import torch.nn.functional as F import torchvision import torchvision.transforms as transforms import torch.optim as optim   from torch.utils.tensorboard import SummaryWriter   batch_size_list = [100, 1000, 10000] lr_list = [.01, .001, .0001, .00001] shuffle = [True,False] def get_num_correct(preds, labels):     return preds.argmax(dim=1).eq(labels).sum().item()   train_set = torchvision.datasets.FashionMNIST(     root='./data/FashionMNIST',     train=True,     download=True,     transform=transforms.Compose([transforms.ToTensor()]) )   # data_loader = torch.utils.data.DataLoader(train_set,batch_size=batch_size,shuffle=True)  # shuffle=True       class Network(nn.Module):     def __init__(self):         super().__init__()         self.conv1 = nn.Conv2d(in_channels=1, out_channels=6, kernel_size=5)         self.conv2 = nn.Conv2d(in_channels=6, out_channels=12, kernel_size=5)           self.fc1 = nn.Linear(in_features=12 * 4 * 4, out_features=120)         self.fc2 = nn.Linear(in_features=120, out_features=60)         self.out = nn.Linear(in_features=60, out_features=10)       def forward(self, t):         # (1) input layer         t = t           # (2) hidden conv layer         t = self.conv1(t)         t = F.relu(t)         t = F.max_pool2d(t, kernel_size=2, stride=2)           # (3) hidden conv layer         t = self.conv2(t)         t = F.relu(t)         t = F.max_pool2d(t, kernel_size=2, stride=2)           # (4) hidden Linear layer         t = t.reshape(-1, 12 * 4 * 4)  # -1表示对行没约束，反正是12*4*4列         t = self.fc1(t)         t = F.relu(t)         # (5) hidden Linear layer         t = self.fc2(t)         t = F.relu(t)         # (6) output layer         t = self.out(t)         # t=F.softmax(t,dim=1) #此处不使用softmax函数，因为在训练中我们使用了交叉熵损失函数，而在torch.nn函数类中，已经在其输入中隐式的         # 执行了一个softmax操作，这里我们只返回最后一个线性变换的结果，也即是 return t，也即意味着我们的网络将使用softmax操作进行训练，但在         # 训练完成后，将不需要额外的计算操纵。           return t   network = Network()   for batch_size in batch_size_list:     for lr in lr_list:         network = Network()           data_loader = torch.utils.data.DataLoader(             train_set, batch_size=batch_size         )         optimizer = optim.Adam(             network.parameters(), lr=lr         )           images, labels = next(iter(data_loader))         grid = torchvision.utils.make_grid(images)           comment=f' batch_size={batch_size} lr={lr}'         tb = SummaryWriter(comment=comment)         tb.add_image('images', grid)         tb.add_graph(network, images)           for epoch in range(5):             total_loss = 0             total_correct = 0             for batch in data_loader:                 images, labels = batch # Get Batch                 preds = network(images) # Pass Batch                 loss = F.cross_entropy(preds, labels) # Calculate Loss                 optimizer.zero_grad() # Zero Gradients                 loss.backward() # Calculate Gradients                 optimizer.step() # Update Weights                   total_loss += loss.item() * batch_size#这里上述用的是mini-batch训练方法，一个batch得loss会被平均，所以乘以size得到总和                 total_correct += get_num_correct(preds, labels)               tb.add_scalar(                 'Loss', total_loss, epoch             )             tb.add_scalar(                 'Number Correct', total_correct, epoch             )             tb.add_scalar(                 'Accuracy', total_correct / len(train_set), epoch             )               for name, param in network.named_parameters():                 tb.add_histogram(name, param, epoch)                 tb.add_histogram(f'{name}.grad', param.grad, epoch)               print(                 "epoch", epoch                 ,"total_correct:", total_correct                 ,"loss:", total_loss             )         tb.close()   f''' from itertools import product   parameters=dict(lr=[.01,.001],batch_size=[10,100,1000],shuffle=[True,False]) # for i,j in parameters.items(): #     print(i,j,sep='\t') para_values=[value for value in parameters.values()] for lr,batch_size,shuffle in product(*para_values):#这里的星号告诉乘积函数把列表中的每个值作为参数，而不是把列表本身作为参数来对待     comment=f' batch_size={batch_size} lr={lr} shuffle={shuffle}'     print(lr,batch_size,shuffle) '''