深度学习与PyTorch入门实战 - 网易云课堂
dragen1860/Deep-Learning-with-PyTorch-Tutorials: 深度学习与PyTorch入门实战视频教程 配套源代码和PPT
主讲：龙良曲

初见PyTorch.pdf

1 深度学习框架简介

image.png

2 PyTorch功能演示

2.1 GPU加速

  1. from pyxllib.xl import PerfTest
  3. import torch
  6. class Process(PerfTest):
  7.     def __init__(self):
  8.         self.a = torch.randn(10000, 1000)
  9.         self.b = torch.randn(1000, 20000)
  11.     def perf_cpu(self):
  12.         return torch.matmul(self.a, self.b).norm(2)
  14.     def perf_gpu(self):
  15.         # device =torch.device('cuda' if torch.cuda.is_available() else 'cpu')
  16.         # torch.device('cpu')、torch.device('cuda:0')
  17.         device = torch.device('cuda')
  18.         return torch.matmul(self.a.to(device), self.b.to(device)).norm(2)
  20.     def perf_gpu2(self):
  21.         # gpu第一次加载比较慢，可以再执行一次看看
  22.         return self.perf_gpu()
  25. print(f'{torch.__version__ = }')
  26. # torch.__version__ = '1.7.1'
  28. Process().perf()
  29. #  cpu 用时(秒) 2.042 运行结果：tensor(414725.4062)
  30. #  gpu 用时(秒) 1.648 运行结果：tensor(447307.6562, device='cuda:0')
  31. # gpu2 用时(秒) 0.029 运行结果：tensor(447307.6562, device='cuda:0')

2.2 自动求导

import torch
from torch import autograd

x = torch.tensor(1.)
a = torch.tensor(1., requires_grad=True)
b = torch.tensor(2., requires_grad=True)
c = torch.tensor(3., requires_grad=True)

y = a ** 2 * x + b * x + c

print('before:', a.grad, b.grad, c.grad)
# before: None None None

grads = autograd.grad(y, [a, b, c])
print('after :', grads[0], grads[1], grads[2])
# after : tensor(2.) tensor(1.) tensor(1.)

image.png

2.3 常用网络层

nn.Linear
nn.Conv2d
nn.LSTM

nn.ReLU
nn.Sigmoid

nn.Softmax
nn.CrossEntropyLoss
nn.MSE