【colab pytorch】使用tensorboardcolab可视化
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 !pip install tensorboardcolab from tensorboardcolab import TensorBoardColab
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((MNIST_MEAN,), (MNIST_STD,)) ])), 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((MNIST_MEAN,), (MNIST_STD,)) ])), 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())) self.tb.save_value('Train Loss', 'train_loss', self.globaliter, loss.item()) 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)) def main(): trainer = Trainer(model_config) for epoch in range(1, trainer.epochs + 1): trainer.train(epoch) trainer.test(epoch) trainer.tb.flush_line('train_loss') if (trainer.save_model): torch.save(trainer.model.state_dict(),"mnist_cnn.pt")
main()
Wait for 8 seconds... TensorBoard link: http://db797eee.ngrok.io 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: 0.013655 Train Epoch: 1 [9600/60000 (16%)] Loss: 0.013620 Train Epoch: 1 [12800/60000 (21%)] Loss: 0.225101 Train Epoch: 1 [16000/60000 (27%)] Loss: 0.248218 Train Epoch: 1 [19200/60000 (32%)] Loss: 0.207354 Train Epoch: 1 [22400/60000 (37%)] Loss: 0.139395 Train Epoch: 1 [25600/60000 (43%)] Loss: 0.206405 Train Epoch: 1 [28800/60000 (48%)] Loss: 0.090241 Train Epoch: 1 [32000/60000 (53%)] Loss: 0.216764 Train Epoch: 1 [35200/60000 (59%)] Loss: 0.295801 Train Epoch: 1 [38400/60000 (64%)] Loss: 0.021000 Train Epoch: 1 [41600/60000 (69%)] Loss: 0.050552 Train Epoch: 1 [44800/60000 (75%)] Loss: 0.238085 Train Epoch: 1 [48000/60000 (80%)] Loss: 0.298676 Train Epoch: 1 [51200/60000 (85%)] Loss: 0.301436 Train Epoch: 1 [54400/60000 (91%)] Loss: 0.271787 Train Epoch: 1 [57600/60000 (96%)] Loss: 0.019811 Test set: Average loss: 0.1088, Accuracy: 9677/10000 (97%) Train Epoch: 2 [0/60000 (0%)] Loss: 0.036418 Train Epoch: 2 [3200/60000 (5%)] Loss: 0.024196 Train Epoch: 2 [6400/60000 (11%)] Loss: 0.029856 Train Epoch: 2 [9600/60000 (16%)] Loss: 0.084013 Train Epoch: 2 [12800/60000 (21%)] Loss: 0.345446 Train Epoch: 2 [16000/60000 (27%)] Loss: 0.453756 Train Epoch: 2 [19200/60000 (32%)] Loss: 0.409682 Train Epoch: 2 [22400/60000 (37%)] Loss: 0.159656 Train Epoch: 2 [25600/60000 (43%)] Loss: 0.009557 Train Epoch: 2 [28800/60000 (48%)] Loss: 0.282826 Train Epoch: 2 [32000/60000 (53%)] Loss: 0.047159 Train Epoch: 2 [35200/60000 (59%)] Loss: 0.379264 Train Epoch: 2 [38400/60000 (64%)] Loss: 0.043181 Train Epoch: 2 [41600/60000 (69%)] Loss: 0.486660 Train Epoch: 2 [44800/60000 (75%)] Loss: 0.108486 Train Epoch: 2 [48000/60000 (80%)] Loss: 0.242821 Train Epoch: 2 [51200/60000 (85%)] Loss: 0.218120 Train Epoch: 2 [54400/60000 (91%)] Loss: 0.381496 Train Epoch: 2 [57600/60000 (96%)] Loss: 0.134828 Test set: Average loss: 0.1861, Accuracy: 9496/10000 (95%) Train Epoch: 3 [0/60000 (0%)] Loss: 0.081437 Train Epoch: 3 [3200/60000 (5%)] Loss: 0.121195 Train Epoch: 3 [6400/60000 (11%)] Loss: 0.054902 Train Epoch: 3 [9600/60000 (16%)] Loss: 0.031254 Train Epoch: 3 [12800/60000 (21%)] Loss: 0.036273 Train Epoch: 3 [16000/60000 (27%)] Loss: 0.162744 Train Epoch: 3 [19200/60000 (32%)] Loss: 0.028073 Train Epoch: 3 [22400/60000 (37%)] Loss: 0.114689 Train Epoch: 3 [25600/60000 (43%)] Loss: 0.139724 Train Epoch: 3 [28800/60000 (48%)] Loss: 0.353534 Train Epoch: 3 [32000/60000 (53%)] Loss: 0.001959 Train Epoch: 3 [35200/60000 (59%)] Loss: 0.117742 Train Epoch: 3 [38400/60000 (64%)] Loss: 0.024078 Train Epoch: 3 [41600/60000 (69%)] Loss: 0.063214 Train Epoch: 3 [44800/60000 (75%)] Loss: 0.068128 Train Epoch: 3 [48000/60000 (80%)] Loss: 0.055476 Train Epoch: 3 [51200/60000 (85%)] Loss: 0.025761 Train Epoch: 3 [54400/60000 (91%)] Loss: 0.490388 Train Epoch: 3 [57600/60000 (96%)] Loss: 0.275244 Test set: Average loss: 0.1570, Accuracy: 9594/10000 (96%) Train Epoch: 4 [0/60000 (0%)] Loss: 0.150237 Train Epoch: 4 [3200/60000 (5%)] Loss: 0.049188 Train Epoch: 4 [6400/60000 (11%)] Loss: 0.008692 Train Epoch: 4 [9600/60000 (16%)] Loss: 0.061360 Train Epoch: 4 [12800/60000 (21%)] Loss: 0.004389 Train Epoch: 4 [16000/60000 (27%)] Loss: 0.027968 Train Epoch: 4 [19200/60000 (32%)] Loss: 0.075881 Train Epoch: 4 [22400/60000 (37%)] Loss: 0.074000 Train Epoch: 4 [25600/60000 (43%)] Loss: 0.069731 Train Epoch: 4 [28800/60000 (48%)] Loss: 0.330368 Train Epoch: 4 [32000/60000 (53%)] Loss: 0.393174 Train Epoch: 4 [35200/60000 (59%)] Loss: 0.318519 Train Epoch: 4 [38400/60000 (64%)] Loss: 0.164669 Train Epoch: 4 [41600/60000 (69%)] Loss: 0.161486 Train Epoch: 4 [44800/60000 (75%)] Loss: 0.017525 Train Epoch: 4 [48000/60000 (80%)] Loss: 0.104918 Train Epoch: 4 [51200/60000 (85%)] Loss: 0.000450 Train Epoch: 4 [54400/60000 (91%)] Loss: 0.128227 Train Epoch: 4 [57600/60000 (96%)] Loss: 0.005374 Test set: Average loss: 0.1227, Accuracy: 9717/10000 (97%)
核心就是标红的地方。
