import torch 
import torch.nn.functional as F 
 
# 方法1，通過定義一個Net類來建立神經(jīng)網(wǎng)絡(luò) 
class Net(torch.nn.Module): 
  def __init__(self, n_feature, n_hidden, n_output): 
    super(Net, self).__init__() 
    self.hidden = torch.nn.Linear(n_feature, n_hidden) 
    self.predict = torch.nn.Linear(n_hidden, n_output) 
 
  def forward(self, x): 
    x = F.relu(self.hidden(x)) 
    x = self.predict(x) 
    return x 
 
net1 = Net(2, 10, 2) 
print('方法1：\n', net1) 
 
# 方法2 通過torch.nn.Sequential快速建立神經(jīng)網(wǎng)絡(luò)結(jié)構(gòu) 
net2 = torch.nn.Sequential( 
  torch.nn.Linear(2, 10), 
  torch.nn.ReLU(), 
  torch.nn.Linear(10, 2), 
  ) 
print('方法2：\n', net2) 
# 經(jīng)驗證，兩種方法構(gòu)建的神經(jīng)網(wǎng)絡(luò)功能相同，結(jié)構(gòu)細節(jié)稍有不同 
 
''''' 
方法1： 
 Net ( 
 (hidden): Linear (2 -> 10) 
 (predict): Linear (10 -> 2) 
) 
方法2： 
 Sequential ( 
 (0): Linear (2 -> 10) 
 (1): ReLU () 
 (2): Linear (10 -> 2) 
) 
''' 

import torch 
from torch.autograd import Variable 
import matplotlib.pyplot as plt 
 
torch.manual_seed(1) # 設(shè)定隨機數(shù)種子 
 
# 創(chuàng)建數(shù)據(jù) 
x = torch.unsqueeze(torch.linspace(-1, 1, 100), dim=1) 
y = x.pow(2) + 0.2*torch.rand(x.size()) 
x, y = Variable(x, requires_grad=False), Variable(y, requires_grad=False) 
 
# 將待保存的神經(jīng)網(wǎng)絡(luò)定義在一個函數(shù)中 
def save(): 
  # 神經(jīng)網(wǎng)絡(luò)結(jié)構(gòu) 
  net1 = torch.nn.Sequential( 
    torch.nn.Linear(1, 10), 
    torch.nn.ReLU(), 
    torch.nn.Linear(10, 1), 
    ) 
  optimizer = torch.optim.SGD(net1.parameters(), lr=0.5) 
  loss_function = torch.nn.MSELoss() 
 
  # 訓(xùn)練部分 
  for i in range(300): 
    prediction = net1(x) 
    loss = loss_function(prediction, y) 
    optimizer.zero_grad() 
    loss.backward() 
    optimizer.step() 
 
  # 繪圖部分 
  plt.figure(1, figsize=(10, 3)) 
  plt.subplot(131) 
  plt.title('net1') 
  plt.scatter(x.data.numpy(), y.data.numpy()) 
  plt.plot(x.data.numpy(), prediction.data.numpy(), 'r-', lw=5) 
 
  # 保存神經(jīng)網(wǎng)絡(luò) 
  torch.save(net1, '7-net.pkl')           # 保存整個神經(jīng)網(wǎng)絡(luò)的結(jié)構(gòu)和模型參數(shù) 
  torch.save(net1.state_dict(), '7-net_params.pkl') # 只保存神經(jīng)網(wǎng)絡(luò)的模型參數(shù) 
 
# 載入整個神經(jīng)網(wǎng)絡(luò)的結(jié)構(gòu)及其模型參數(shù) 
def reload_net(): 
  net2 = torch.load('7-net.pkl') 
  prediction = net2(x) 
 
  plt.subplot(132) 
  plt.title('net2') 
  plt.scatter(x.data.numpy(), y.data.numpy()) 
  plt.plot(x.data.numpy(), prediction.data.numpy(), 'r-', lw=5) 
 
# 只載入神經(jīng)網(wǎng)絡(luò)的模型參數(shù)，神經(jīng)網(wǎng)絡(luò)的結(jié)構(gòu)需要與保存的神經(jīng)網(wǎng)絡(luò)相同的結(jié)構(gòu) 
def reload_params(): 
  # 首先搭建相同的神經(jīng)網(wǎng)絡(luò)結(jié)構(gòu) 
  net3 = torch.nn.Sequential( 
    torch.nn.Linear(1, 10), 
    torch.nn.ReLU(), 
    torch.nn.Linear(10, 1), 
    ) 
 
  # 載入神經(jīng)網(wǎng)絡(luò)的模型參數(shù) 
  net3.load_state_dict(torch.load('7-net_params.pkl')) 
  prediction = net3(x) 
 
  plt.subplot(133) 
  plt.title('net3') 
  plt.scatter(x.data.numpy(), y.data.numpy()) 
  plt.plot(x.data.numpy(), prediction.data.numpy(), 'r-', lw=5) 
 
# 運行測試 
save() 
reload_net() 
reload_params()