代码语言：javascript复制
import datetime
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torch.utils.data import Dataset, DataLoader
from torchvision import transforms, utils, datasets
from tensorflow import summary%load_ext tensorboard根据情况换成
%load_ext tensorboard.notebook
代码语言：javascript复制class Network(nn.Module):
def __init__(self):
super(Network, self).__init__()
self.conv1 = nn.Conv2d(1, 20, 5, 1)
self.conv2 = nn.Conv2d(20, 50, 5, 1)
self.fc1 = nn.Linear(4*4*50, 500)
self.fc2 = nn.Linear(500, 10)
def forward(self, x):
x = F.relu(self.conv1(x))
x = F.max_pool2d(x, 2, 2)
x = F.relu(self.conv2(x))
x = F.max_pool2d(x, 2, 2)
x = x.view(-1, 4*4*50)
x = F.relu(self.fc1(x))
x = self.fc2(x)
return F.log_softmax(x, dim=1)
class Config:
def __init__(self, **kwargs):
for key, value in kwargs.items():
setattr(self, key, value)
model_config = Config(
cuda = True if torch.cuda.is_available() else False,
device = torch.device("cuda" if torch.cuda.is_available() else "cpu"),
seed = 2,
lr = 0.01,
epochs = 4,
save_model = False,
batch_size = 32,
log_interval = 100
)
class Trainer:
def __init__(self, config):
self.cuda = config.cuda
self.device = config.device
self.seed = config.seed
self.lr = config.lr
self.epochs = config.epochs
self.save_model = config.save_model
self.batch_size = config.batch_size
self.log_interval = config.log_interval
self.globaliter = 0
#self.tb = TensorBoardColab()
torch.manual_seed(self.seed)
kwargs = {'num_workers': 1, 'pin_memory': True} if self.cuda else {}
self.train_loader = torch.utils.data.DataLoader(
datasets.MNIST('../data', train=True, download=True,
transform=transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.1307,), (0.3081,))
])),
batch_size=self.batch_size, shuffle=True, **kwargs)
self.test_loader = torch.utils.data.DataLoader(
datasets.MNIST('../data', train=False, transform=transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.1307,), (0.3081,))
])),
batch_size=self.batch_size, shuffle=True, **kwargs)
self.model = Network().to(self.device)
self.optimizer = optim.Adam(self.model.parameters(), lr=self.lr)
def train(self, epoch):
self.model.train()
for batch_idx, (data, target) in enumerate(self.train_loader):
self.globaliter = 1
data, target = data.to(self.device), target.to(self.device)
self.optimizer.zero_grad()
predictions = self.model(data)
loss = F.nll_loss(predictions, target)
loss.backward()
self.optimizer.step()
if batch_idx % self.log_interval == 0:
print('Train Epoch: {} [{}/{} ({:.0f}%)]tLoss: {:.6f}'.format(
epoch, batch_idx * len(data), len(self.train_loader.dataset),
100. * batch_idx / len(self.train_loader), loss.item()))
with train_summary_writer.as_default():
summary.scalar('loss', loss.item(), step=self.globaliter)
def test(self, epoch):
self.model.eval()
test_loss = 0
correct = 0
with torch.no_grad():
for data, target in self.test_loader:
data, target = data.to(self.device), target.to(self.device)
predictions = self.model(data)
test_loss = F.nll_loss(predictions, target, reduction='sum').item()
prediction = predictions.argmax(dim=1, keepdim=True)
correct = prediction.eq(target.view_as(prediction)).sum().item()
test_loss /= len(self.test_loader.dataset)
accuracy = 100. * correct / len(self.test_loader.dataset)
print('nTest set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)n'.format(
test_loss, correct, len(self.test_loader.dataset), accuracy))
with test_summary_writer.as_default():
summary.scalar('loss', test_loss, step=self.globaliter)
summary.scalar('accuracy', accuracy, step=self.globaliter)
def main():
trainer = Trainer(model_config)
for epoch in range(1, trainer.epochs 1):
trainer.train(epoch)
trainer.test(epoch)
if (trainer.save_model):
torch.save(trainer.model.state_dict(),"mnist_cnn.pt")current_time = str(datetime.datetime.now().timestamp())
train_log_dir = 'logs/tensorboard/train/' current_time
test_log_dir = 'logs/tensorboard/test/' current_time
train_summary_writer = summary.create_file_writer(train_log_dir)
test_summary_writer = summary.create_file_writer(test_log_dir)%tensorboard --logdir logs/tensorboardmain()Train Epoch: 1 [0/60000 (0%)] Loss: 2.320306
Train Epoch: 1 [3200/60000 (5%)] Loss: 0.881239
Train Epoch: 1 [6400/60000 (11%)] Loss: 0.014427
Train Epoch: 1 [9600/60000 (16%)] Loss: 0.046511
Train Epoch: 1 [12800/60000 (21%)] Loss: 0.194090
Train Epoch: 1 [16000/60000 (27%)] Loss: 0.178779
Train Epoch: 1 [19200/60000 (32%)] Loss: 0.437568
Train Epoch: 1 [22400/60000 (37%)] Loss: 0.058614
Train Epoch: 1 [25600/60000 (43%)] Loss: 0.051354
Train Epoch: 1 [28800/60000 (48%)] Loss: 0.339627
Train Epoch: 1 [32000/60000 (53%)] Loss: 0.057814
Train Epoch: 1 [35200/60000 (59%)] Loss: 0.216959
Train Epoch: 1 [38400/60000 (64%)] Loss: 0.111091
Train Epoch: 1 [41600/60000 (69%)] Loss: 0.268371
Train Epoch: 1 [44800/60000 (75%)] Loss: 0.129569
Train Epoch: 1 [48000/60000 (80%)] Loss: 0.392319
Train Epoch: 1 [51200/60000 (85%)] Loss: 0.374106
Train Epoch: 1 [54400/60000 (91%)] Loss: 0.145877
Train Epoch: 1 [57600/60000 (96%)] Loss: 0.136342
Test set: Average loss: 0.1660, Accuracy: 9497/10000 (95%)
Train Epoch: 2 [0/60000 (0%)] Loss: 0.215095
Train Epoch: 2 [3200/60000 (5%)] Loss: 0.064202
Train Epoch: 2 [6400/60000 (11%)] Loss: 0.059504
Train Epoch: 2 [9600/60000 (16%)] Loss: 0.116854
Train Epoch: 2 [12800/60000 (21%)] Loss: 0.259310
Train Epoch: 2 [16000/60000 (27%)] Loss: 0.280154
Train Epoch: 2 [19200/60000 (32%)] Loss: 0.260245
Train Epoch: 2 [22400/60000 (37%)] Loss: 0.039311
Train Epoch: 2 [25600/60000 (43%)] Loss: 0.049329
Train Epoch: 2 [28800/60000 (48%)] Loss: 0.437081
Train Epoch: 2 [32000/60000 (53%)] Loss: 0.094939
Train Epoch: 2 [35200/60000 (59%)] Loss: 0.311777
Train Epoch: 2 [38400/60000 (64%)] Loss: 0.076921
Train Epoch: 2 [41600/60000 (69%)] Loss: 0.800094
Train Epoch: 2 [44800/60000 (75%)] Loss: 0.074938
Train Epoch: 2 [48000/60000 (80%)] Loss: 0.240811
Train Epoch: 2 [51200/60000 (85%)] Loss: 0.303044
Train Epoch: 2 [54400/60000 (91%)] Loss: 0.372847
Train Epoch: 2 [57600/60000 (96%)] Loss: 0.290946
Test set: Average loss: 0.1341, Accuracy: 9634/10000 (96%)
Train Epoch: 3 [0/60000 (0%)] Loss: 0.092767
Train Epoch: 3 [3200/60000 (5%)] Loss: 0.038457
Train Epoch: 3 [6400/60000 (11%)] Loss: 0.005179
Train Epoch: 3 [9600/60000 (16%)] Loss: 0.168411
Train Epoch: 3 [12800/60000 (21%)] Loss: 0.171331
Train Epoch: 3 [16000/60000 (27%)] Loss: 0.267252
Train Epoch: 3 [19200/60000 (32%)] Loss: 0.072991
Train Epoch: 3 [22400/60000 (37%)] Loss: 0.034315
Train Epoch: 3 [25600/60000 (43%)] Loss: 0.143128
Train Epoch: 3 [28800/60000 (48%)] Loss: 0.324783
Train Epoch: 3 [32000/60000 (53%)] Loss: 0.049743
Train Epoch: 3 [35200/60000 (59%)] Loss: 0.090172
Train Epoch: 3 [38400/60000 (64%)] Loss: 0.002107
Train Epoch: 3 [41600/60000 (69%)] Loss: 0.025945
Train Epoch: 3 [44800/60000 (75%)] Loss: 0.054859
Train Epoch: 3 [48000/60000 (80%)] Loss: 0.009291
Train Epoch: 3 [51200/60000 (85%)] Loss: 0.010495
Train Epoch: 3 [54400/60000 (91%)] Loss: 0.132548
Train Epoch: 3 [57600/60000 (96%)] Loss: 0.005778
Test set: Average loss: 0.1570, Accuracy: 9553/10000 (96%)
Train Epoch: 4 [0/60000 (0%)] Loss: 0.103177
Train Epoch: 4 [3200/60000 (5%)] Loss: 0.087844
Train Epoch: 4 [6400/60000 (11%)] Loss: 0.066604
Train Epoch: 4 [9600/60000 (16%)] Loss: 0.052869
Train Epoch: 4 [12800/60000 (21%)] Loss: 0.091576
Train Epoch: 4 [16000/60000 (27%)] Loss: 0.094903
Train Epoch: 4 [19200/60000 (32%)] Loss: 0.247008
Train Epoch: 4 [22400/60000 (37%)] Loss: 0.037751
Train Epoch: 4 [25600/60000 (43%)] Loss: 0.067071
Train Epoch: 4 [28800/60000 (48%)] Loss: 0.191988
Train Epoch: 4 [32000/60000 (53%)] Loss: 0.403029
Train Epoch: 4 [35200/60000 (59%)] Loss: 0.547171
Train Epoch: 4 [38400/60000 (64%)] Loss: 0.187923
Train Epoch: 4 [41600/60000 (69%)] Loss: 0.231193
Train Epoch: 4 [44800/60000 (75%)] Loss: 0.010785
Train Epoch: 4 [48000/60000 (80%)] Loss: 0.077892
Train Epoch: 4 [51200/60000 (85%)] Loss: 0.093144
Train Epoch: 4 [54400/60000 (91%)] Loss: 0.004715
Train Epoch: 4 [57600/60000 (96%)] Loss: 0.083726
Test set: Average loss: 0.1932, Accuracy: 9584/10000 (96%)核心就是标红的地方。
