深度学习04 -模型管理

访问模型参数

import torch
from torch.utils import data
from torch import nn

net = nn.Sequential(nn.Linear(2, 4), nn.ReLU(), nn.Linear(4, 1))

print((net[0].state_dict()))
print(net[0].bias)
print(net[0].bias.data)

rderedDict([('weight', tensor([[ 0.2167, -0.4856],
        [-0.6567, -0.4602],
        [ 0.1185,  0.1153],
        [-0.0756, -0.6647]])), ('bias', tensor([0.4804, 0.4804, 0.3834, 0.7041]))])
Parameter containing:
tensor([0.4804, 0.4804, 0.3834, 0.7041], requires_grad=True)
tensor([0.4804, 0.4804, 0.3834, 0.7041])

保存Tensor数据

>>> x = torch.arange(4)
>>> torch.save(x, 'x-file')
>>> x2 = torch.load('x-file')
>>> x2
tensor([0, 1, 2, 3])

保存模型参数

class MLP(nn.Module):
    def __init__(self):
        super().__init__()
        self.hidden = nn.Linear(1, 2)
        self.output = nn.Linear(2, 2)

    def forward(self, x):
        return self.output(F.relu(self.hidden(x)))

net = MLP()

torch.save(net.state_dict(), 'mlp.params')

clone = MLP()
clone.load_state_dict(torch.load('mlp.params'))
clone.eval() # 从train模式调整为test模式，不再进行训练更改梯度

net.state_dict() , clone.state_dict()

(OrderedDict([('hidden.weight', tensor([[-0.6865],
                       [-0.2476]])),
              ('hidden.bias', tensor([ 0.3468, -0.3090])),
              ('output.weight', tensor([[ 0.1645, -0.1768],
                       [ 0.7000, -0.6238]])),
              ('output.bias', tensor([-0.0137, -0.4359]))]),
 OrderedDict([('hidden.weight', tensor([[-0.6865],
                       [-0.2476]])),
              ('hidden.bias', tensor([ 0.3468, -0.3090])),
              ('output.weight', tensor([[ 0.1645, -0.1768],
                       [ 0.7000, -0.6238]])),
              ('output.bias', tensor([-0.0137, -0.4359]))]))