Weight&Biases教程

简介

Weights & Biases 可以帮助跟踪你的机器学习项目。使用工具记录运行中的超参数和输出指标(Metric)，然后对结果进行可视化和比较，并快速与同事分享你的发现。

好处

首先它和tensorboard的功能类似，但是区别在于可以通过真正的网址随时进行访问，同时他人也能看到你的效果，省去了麻烦的步骤，同时还能方便自己随时随地监工自己的网络运行情况，同时它的sweeps能自己调参优化模型

工具
仪表盘：跟踪实验、可视化结果。
报告：保存和分享可复制的成果/结论。
Sweeps：通过调节超参数来优化模型
Artifacts : 数据集和模型版本化，流水线跟踪。

网址

https://wandb.ai/site,注册一个账号，个人是免费的，然后创建一个项目，然后按照官网提示安装pip install wandb

测试demo

import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torchvision import datasets, transforms
import wandb


batch_size = 200
learning_rate = 0.01
epochs = 30

wandb.init(
    # Set the project where this run will be logged
    project=" ",#写自己的
    entity=" ",#写自己的
    # Track hyperparameters and run metadata
    config={
   
        "learning_rate": learning_rate,
        "architecture": "MLP",
        "dataset": "MNIST",
        "epochs": epochs,
    })

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=batch_size, shuffle=True)
test_loader = torch.utils.data.DataLoader(
    datasets.MNIST('../data', train=False, transform=transforms.Compose([
        transforms.ToTensor(),
        # transforms.Normalize((0.1307,), (0.3081,))
    ])),
    batch_size=batch_size, shuffle=True)


class MLP(nn.Module):

    def __init__(self):
        super(MLP, self).__init__()

        self.model = nn.Sequential(
            nn.Linear(784, 200),
            nn.LeakyReLU(inplace=True),
            nn.Linear(200, 200),
            nn.LeakyReLU(inplace=True),
            nn.Linear(200, 10),
            nn.LeakyReLU(inplace=True),
        )

    def forward(self, x):
        x = self.model(x)

        return x


device = torch.device('cuda:0')
net = MLP().to(device)
optimizer = optim.SGD(net.parameters(), lr=learning_rate)
criteon = nn.CrossEntropyLoss().to(device)


global_step = 0

for epoch in range(epochs):
    net.train()
    for batch_idx, (data, target) in enumerate(train_loader):
        data = data.view(-1, 28 * 28)
        data, target = data.to(device), target.cuda()

        logits = net(data)
        loss = criteon(logits, target)

        optimizer.zero_grad()
        loss.backward()
        # print(w1.grad.norm(), w2.grad.norm())
        optimizer.step()

        global_step += 1

        if batch_idx % 100 == 0:
            print('Train Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.6f}'.format(
                epoch, batch_idx * len(data), len(train_loader.dataset),
                       100. * batch_idx / len(train_loader), loss.item()))
            #展示数据
            wandb.log({
   "Train loss": loss.item()})
    net.eval()
    test_loss = 0
    correct = 0
    for data, target in test_loader:
        data = data.view(-1, 28 * 28)
        data, target = data.to(device), target.cuda()
        logits = net(data)
        test_loss += criteon(logits, target).item()

        pred = logits.argmax(dim=1)
        correct += pred.eq(target).float().sum().item()

    test_loss /= len(test_loader.dataset)
    print('\nTest set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n'