pytorch使用說明2

網絡保存與加載

1.保存

torch.manual_seed(1)    # reproducible

# 假數據
x = torch.unsqueeze(torch.linspace(-1, 1, 100), dim=1)  # x data (tensor), shape=(100, 1)
y = x.pow(2) + 0.2*torch.rand(x.size())  # noisy y data (tensor), shape=(100, 1)

def save():
    # 建網絡
    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_func = torch.nn.MSELoss()

    # 訓練
    for t in range(100):
        prediction = net1(x)
        loss = loss_func(prediction, y)
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()
        
torch.save(net1, 'net.pkl')  # 保存整個網絡
torch.save(net.state_dict(), 'net_params.pkl') # 只保存網絡中的參數(速度快，佔內存少)

2.加載網絡

def restore_net():
    # restore entire net1 to net2
    net2 = torch.load('net.pkl')
    prediction = net2(x)
    

# 只提取網絡參數
def restore_params():
    # 新建 net3
    net3 = torch.nn.Sequential(
        torch.nn.Linear(1, 10),
        torch.nn.ReLU(),
        torch.nn.Linear(10, 1)
    )

    # 將保存的參數複製到 net3
    net3.load_state_dict(torch.load('net_params.pkl'))
    prediction = net3(x)
   

# 保存 net1 (1. 整個網絡, 2. 只有參數)
save()
# 提取整個網絡
restore_net()
# 提取網絡參數, 複製到新網絡
restore_params()

3.批訓練

DataLoader是torch給你用來包裝你的數據的工具。因此要將本身的(numpy array或其餘)數據形式轉換成Tensor, 而後再放進這個包裝器中。使用DataLoader的好處就是幫你有效地迭代數據。

import torch
import torch.utils.data as Data
torch.manual_seed(1)    # reproducible

BATCH_SIZE = 5      # 批訓練的數據個數

x = torch.linspace(1, 10, 10)       # x data (torch tensor)
y = torch.linspace(10, 1, 10)       # y data (torch tensor)

# 先轉換成 torch 能識別的 Dataset
torch_dataset = Data.TensorDataset(data_tensor=x, target_tensor=y)

# 把 dataset 放入 DataLoader
loader = Data.DataLoader(
    dataset=torch_dataset,      # torch TensorDataset format
    batch_size=BATCH_SIZE,      # mini batch size
    shuffle=True,               # 要不要打亂數據 (打亂比較好)
    num_workers=2,              # 多線程來讀數據
)

for epoch in range(3):   # 訓練全部!整套!數據 3 次
    for step, (batch_x, batch_y) in enumerate(loader):  # 每一步 loader 釋放一小批數據用來學習
        # 假設這裏就是你訓練的地方...

        # 打出來一些數據
        print('Epoch: ', epoch, '| Step: ', step, '| batch x: ',
              batch_x.numpy(), '| batch y: ', batch_y.numpy())

"""
Epoch:  0 | Step:  0 | batch x:  [ 6.  7.  2.  3.  1.] | batch y:  [  5.   4.   9.   8.  10.]
Epoch:  0 | Step:  1 | batch x:  [  9.  10.   4.   8.   5.] | batch y:  [ 2.  1.  7.  3.  6.]
Epoch:  1 | Step:  0 | batch x:  [  3.   4.   2.   9.  10.] | batch y:  [ 8.  7.  9.  2.  1.]
Epoch:  1 | Step:  1 | batch x:  [ 1.  7.  8.  5.  6.] | batch y:  [ 10.   4.   3.   6.   5.]
Epoch:  2 | Step:  0 | batch x:  [ 3.  9.  2.  6.  7.] | batch y:  [ 8.  2.  9.  5.  4.]
Epoch:  2 | Step:  1 | batch x:  [ 10.   4.   8.   1.   5.] | batch y:  [  1.   7.   3.  10.   6.]
"""

當數據最後不足batch時，就會返回這個epoch中剩下的數據。