474 KiB
474 KiB
In [2]:
import torch
import d2l
import numpy
import torch.nn as nn
import torch.nn.functional as FIn [3]:
net = nn.Sequential(nn.Linear(20, 256), nn.ReLU(), nn.Linear(256, 10))
X = torch.rand(2, 20)
net(X)Out [3]:
tensor([[ 0.0041, -0.3465, -0.2096, 0.2304, -0.1043, 0.0066, 0.1817, 0.0355,
0.2685, -0.0461],
[-0.0932, -0.1621, -0.1244, 0.2398, -0.0759, 0.0680, 0.1511, 0.0224,
0.2522, -0.0228]], grad_fn=<AddmmBackward0>)In [4]:
class MLP(nn.Module):
def __init__(self):
super().__init__()
self.hidden=nn.Linear(20,256)
self.out=nn.Linear(256,10)
def forward(self,X):
return self.out(F.relu(self.hidden(X)))
In [5]:
net=MLP()
net(X)Out [5]:
tensor([[-0.2165, 0.1394, 0.0867, 0.0692, 0.2914, -0.1427, 0.2218, -0.0533,
-0.2137, 0.0044],
[-0.2020, 0.0648, 0.0514, 0.0500, 0.2555, -0.1679, 0.1621, -0.1462,
-0.2527, 0.0386]], grad_fn=<AddmmBackward0>)In [6]:
class FixedHiddenMLP(nn.Module):
def __init__(self):
super().__init__()
# 不计算梯度的随机权重参数。因此其在训练期间保持不变
self.rand_weight = torch.rand((20, 20), requires_grad=False)
self.linear = nn.Linear(20, 20)
def forward(self, X):
X = self.linear(X)
# 使用创建的常量参数以及relu和mm函数
X = F.relu(torch.mm(X, self.rand_weight) + 1)
# 复用全连接层。这相当于两个全连接层共享参数
X = self.linear(X)
# 控制流
while X.abs().sum() > 1:
X /= 2
return X.sum()In [7]:
net = FixedHiddenMLP()
net(X)Out [7]:
tensor(-0.0023, grad_fn=<SumBackward0>)
In [8]:
class NestMLP(nn.Module):
def __init__(self):
super().__init__()
self.net = nn.Sequential(nn.Linear(20, 64), nn.ReLU(),
nn.Linear(64, 32), nn.ReLU())
self.linear = nn.Linear(32, 16)
def forward(self, X):
return self.linear(self.net(X))
chimera = nn.Sequential(NestMLP(), nn.Linear(16, 20), FixedHiddenMLP())
chimera(X)In [9]:
net = nn.Sequential(nn.Linear(4, 8), nn.ReLU(), nn.Linear(8, 1))
X = torch.rand(size=(2, 4))
net(X)Out [9]:
tensor([[-0.1265],
[-0.0471]], grad_fn=<AddmmBackward0>)In [10]:
print(net[2].state_dict())OrderedDict([('weight', tensor([[ 0.0136, -0.1015, 0.1191, 0.2722, 0.3456, -0.0650, -0.0437, -0.2806]])), ('bias', tensor([-0.0945]))])
In [11]:
net[2].state_dict()Out [11]:
OrderedDict([('weight',
tensor([[ 0.0136, -0.1015, 0.1191, 0.2722, 0.3456, -0.0650, -0.0437, -0.2806]])),
('bias', tensor([-0.0945]))])In [12]:
print(type(net[2].bias))<class 'torch.nn.parameter.Parameter'>
In [13]:
print(net[2].bias)
print(net[2].bias.data)
Parameter containing: tensor([-0.0945], requires_grad=True) tensor([-0.0945])
In [14]:
net[2].weight.grad==NoneOut [14]:
True
In [15]:
print(*[(name, param.shape) for name, param in net[0].named_parameters()])
print(*[(name, param.shape) for name, param in net.named_parameters()])('weight', torch.Size([8, 4])) ('bias', torch.Size([8]))
('0.weight', torch.Size([8, 4])) ('0.bias', torch.Size([8])) ('2.weight', torch.Size([1, 8])) ('2.bias', torch.Size([1]))
In [16]:
net.state_dict()['2.bias'].dataOut [16]:
tensor([-0.0945])
In [17]:
def block1():
return nn.Sequential(nn.Linear(4, 8), nn.ReLU(), nn.Linear(8, 4),nn.ReLU())
def block2():
net = nn.Sequential()
for i in range(4):
net.add_module(f'block{i}', block1())
return netIn [18]:
rgnet = nn.Sequential(block2(),nn.Linear(4,1))
rgnet(X)Out [18]:
tensor([[0.0117],
[0.0117]], grad_fn=<AddmmBackward0>)In [19]:
print(rgnet)Sequential(
(0): Sequential(
(block0): Sequential(
(0): Linear(in_features=4, out_features=8, bias=True)
(1): ReLU()
(2): Linear(in_features=8, out_features=4, bias=True)
(3): ReLU()
)
(block1): Sequential(
(0): Linear(in_features=4, out_features=8, bias=True)
(1): ReLU()
(2): Linear(in_features=8, out_features=4, bias=True)
(3): ReLU()
)
(block2): Sequential(
(0): Linear(in_features=4, out_features=8, bias=True)
(1): ReLU()
(2): Linear(in_features=8, out_features=4, bias=True)
(3): ReLU()
)
(block3): Sequential(
(0): Linear(in_features=4, out_features=8, bias=True)
(1): ReLU()
(2): Linear(in_features=8, out_features=4, bias=True)
(3): ReLU()
)
)
(1): Linear(in_features=4, out_features=1, bias=True)
)
In [20]:
rgnet[0][1][0].bias.dataOut [20]:
tensor([ 0.2396, -0.2293, -0.3365, 0.0070, -0.0166, -0.2328, -0.1627, 0.3407])
In [21]:
def init_normal(m):
if type(m) == nn.Linear:
nn.init.normal_(m.weight, mean=0, std=0.01)
nn.init.zeros_(m.bias)
net.apply(init_normal)
net[0].weight.data[0], net[0].bias.data[0]Out [21]:
(tensor([ 0.0166, 0.0092, 0.0013, -0.0031]), tensor(0.))
In [22]:
def init_xavier(m):
if type(m) == nn.Linear:
nn.init.xavier_uniform_(m.weight)
def init_42(m):
if type(m) == nn.Linear:
nn.init.constant_(m.weight, 42)
net[0].apply(init_xavier)
net[2].apply(init_42)
print(net[0].weight.data[0])
print(net[2].weight.data)tensor([-0.2085, 0.4344, -0.3960, 0.5868]) tensor([[42., 42., 42., 42., 42., 42., 42., 42.]])
In [23]:
x = torch.arange(4)
torch.save(x, 'x-file')In [24]:
x2 = torch.load('x-file')
x2Out [24]:
tensor([0, 1, 2, 3])
In [25]:
class MLP(nn.Module):
def __init__(self):
super().__init__()
self.hidden = nn.Linear(20, 256)
self.output = nn.Linear(256, 10)
def forward(self, x):
return self.output(F.relu(self.hidden(x)))
net = MLP()
X = torch.randn(size=(2, 20))
Y = net(X)In [26]:
torch.save(net.state_dict(), 'mlp.params')In [27]:
clone = MLP()
clone.load_state_dict(torch.load('mlp.params'))
clone.eval()Out [27]:
MLP( (hidden): Linear(in_features=20, out_features=256, bias=True) (output): Linear(in_features=256, out_features=10, bias=True) )
In [28]:
Y_clone = clone(X)
Y_clone == YOut [28]:
tensor([[True, True, True, True, True, True, True, True, True, True],
[True, True, True, True, True, True, True, True, True, True]])In [29]:
def corr2d(X,K):
h,w=K.shape
Y=torch.ones((X.shape[0]-h+1,X.shape[1]-w+1))
for i in range(Y.shape[0]):
for j in range(Y.shape[1]):
Y[i,j]=(X[i:i+h,j:j+w]*K).sum()
return Y
In [30]:
X = torch.tensor([[0.0, 1.0, 2.0], [3.0, 4.0, 5.0], [6.0, 7.0, 8.0]])
K = torch.tensor([[0.0, 1.0], [2.0, 3.0]])
corr2d(X,K)Out [30]:
tensor([[19., 25.],
[37., 43.]])In [31]:
class Conv2D(nn.Module):
def __init__(self, kernel_size):
super().__init__()
self.weight = nn.Parameter(torch.rand(kernel_size))
self.bias = nn.Parameter(torch.zeros(1))
def forward(self, x):
return corr2d(x, self.weight) + self.bias
In [32]:
X = torch.ones((6, 8))
X[:, 2:6] = 0
XOut [32]:
tensor([[1., 1., 0., 0., 0., 0., 1., 1.],
[1., 1., 0., 0., 0., 0., 1., 1.],
[1., 1., 0., 0., 0., 0., 1., 1.],
[1., 1., 0., 0., 0., 0., 1., 1.],
[1., 1., 0., 0., 0., 0., 1., 1.],
[1., 1., 0., 0., 0., 0., 1., 1.]])In [33]:
K = torch.tensor([[1.0, -1.0]])
Y = corr2d(X, K)
YOut [33]:
tensor([[ 0., 1., 0., 0., 0., -1., 0.],
[ 0., 1., 0., 0., 0., -1., 0.],
[ 0., 1., 0., 0., 0., -1., 0.],
[ 0., 1., 0., 0., 0., -1., 0.],
[ 0., 1., 0., 0., 0., -1., 0.],
[ 0., 1., 0., 0., 0., -1., 0.]])In [34]:
corr2d(X.t(), K)Out [34]:
tensor([[0., 0., 0., 0., 0.],
[0., 0., 0., 0., 0.],
[0., 0., 0., 0., 0.],
[0., 0., 0., 0., 0.],
[0., 0., 0., 0., 0.],
[0., 0., 0., 0., 0.],
[0., 0., 0., 0., 0.],
[0., 0., 0., 0., 0.]])In [35]:
conv2d = nn.Conv2d(1,1, kernel_size=(1, 2), bias=False)In [36]:
X = X.reshape((1, 1, 6, 8))
Y = Y.reshape((1, 1, 6, 7))
lr = 3e-2In [37]:
for i in range(100):
Y_hat = conv2d(X)
l = (Y_hat - Y) ** 2
conv2d.zero_grad()
l.sum().backward()
# 迭代卷积核
conv2d.weight.data[:] -= lr * conv2d.weight.grad
if (i + 1) % 20 == 0:
print(f'epoch {i+1}, loss {l.sum():.3f}')epoch 20, loss 0.000 epoch 40, loss 0.000 epoch 60, loss 0.000 epoch 80, loss 0.000 epoch 100, loss 0.000
In [38]:
conv2d.weight.data.reshape((1, 2))
Out [38]:
tensor([[ 1.0000, -1.0000]])
In [39]:
# 为了方便起见，我们定义了一个计算卷积层的函数。
# 此函数初始化卷积层权重，并对输入和输出提高和缩减相应的维数
def comp_conv2d(conv2d, X):
# 这里的（1，1）表示批量大小和通道数都是1
X = X.reshape((1, 1) + X.shape)
Y = conv2d(X)
# 省略前两个维度：批量大小和通道
return Y.reshape(Y.shape[2:])
# 请注意，这里每边都填充了1行或1列，因此总共添加了2行或2列
conv2d = nn.Conv2d(1, 1, kernel_size=3, padding=1)In [40]:
X = torch.rand(size=(8, 8))
comp_conv2d(conv2d, X).shapeOut [40]:
torch.Size([8, 8])
In [41]:
conv2d = nn.Conv2d(1, 1, kernel_size=(5, 3), padding=(2, 1))
comp_conv2d(conv2d, X).shapeOut [41]:
torch.Size([8, 8])
In [42]:
conv2d = nn.Conv2d(1, 1, kernel_size=3, padding=1, stride=2)
comp_conv2d(conv2d, X).shapeOut [42]:
torch.Size([4, 4])
In [43]:
conv2d = nn.Conv2d(1, 1, kernel_size=(3, 5), padding=(0, 1), stride=(3, 4))
comp_conv2d(conv2d, X).shapeOut [43]:
torch.Size([2, 2])
In [44]:
def corr2d_multi_in(X,K):
return sum(corr2d(x,k) for x,k in zip(X,K))
X = torch.tensor([[[0.0, 1.0, 2.0], [3.0, 4.0, 5.0], [6.0, 7.0, 8.0]],
[[1.0, 2.0, 3.0], [4.0, 5.0, 6.0], [7.0, 8.0, 9.0]]])
K = torch.tensor([[[0.0, 1.0], [2.0, 3.0]], [[1.0, 2.0], [3.0, 4.0]]])
corr2d_multi_in(X, K)Out [44]:
tensor([[ 56., 72.],
[104., 120.]])In [45]:
def corr2d_multi_in_out(X,K) ->torch.Tensor :
return torch.stack([corr2d_multi_in(X,k) for k in K],0)
In [46]:
K = torch.stack((K, K + 1, K + 2), 0)
K.shapeOut [46]:
torch.Size([3, 2, 2, 2])
In [47]:
corr2d_multi_in_out(X, K)Out [47]:
tensor([[[ 56., 72.],
[104., 120.]],
[[ 76., 100.],
[148., 172.]],
[[ 96., 128.],
[192., 224.]]])In [48]:
def corr2d_multi_in_out_1x1(X, K):
h_i,h,w=X.shape
h_o=K.shape[0]
X=X.reshape((h_i,h*w))
print(X.shape)
K=K.reshape((h_o,h_i))
print(K.shape)
Y=torch.matmul(K,X)
return Y.reshape((h_o,h,w))In [49]:
X = torch.normal(0, 1, (3, 3, 3))
K = torch.normal(0, 1, (2, 3, 1, 1))In [50]:
Y1 = corr2d_multi_in_out_1x1(X, K)torch.Size([3, 9]) torch.Size([2, 3])
In [51]:
Y2 = corr2d_multi_in_out(X, K)
assert float(torch.abs(Y1 - Y2).sum()) < 1e-6In [52]:
def pool2d(X,pool_size,mode='max'):
p_h,p_w =pool_size
Y = torch.zeros((X.shape[0]-p_h+1,X.shape[1]-p_w+1))
for i in range(Y.shape[0]):
for j in range(Y.shape[1]):
match mode:
case 'max':
Y[i,j]=X[i:i+p_h,j:j+p_w].max()
case 'avg':
Y[i,j]=X[i:i+p_h,j:j+p_w].mean()
return YIn [53]:
X = torch.tensor([[0.0, 1.0, 2.0], [3.0, 4.0, 5.0], [6.0, 7.0, 8.0]])
pool2d(X, (2, 2))Out [53]:
tensor([[4., 5.],
[7., 8.]])In [54]:
pool2d(X, (2, 2), 'avg')Out [54]:
tensor([[2., 3.],
[5., 6.]])In [55]:
X = torch.arange(16, dtype=torch.float32).reshape((1, 1, 4, 4))
XOut [55]:
tensor([[[[ 0., 1., 2., 3.],
[ 4., 5., 6., 7.],
[ 8., 9., 10., 11.],
[12., 13., 14., 15.]]]])In [56]:
pool2d=nn.MaxPool2d(3)
pool2d(X)Out [56]:
tensor([[[[10.]]]])
In [57]:
pool2d = nn.MaxPool2d(3, padding=1, stride=2)
pool2d(X)Out [57]:
tensor([[[[ 5., 7.],
[13., 15.]]]])In [58]:
pool2d = nn.MaxPool2d((2, 3), stride=(2, 3), padding=(0, 1))
pool2d(X)Out [58]:
tensor([[[[ 5., 7.],
[13., 15.]]]])In [59]:
X = torch.cat((X, X + 1), 1)
XOut [59]:
tensor([[[[ 0., 1., 2., 3.],
[ 4., 5., 6., 7.],
[ 8., 9., 10., 11.],
[12., 13., 14., 15.]],
[[ 1., 2., 3., 4.],
[ 5., 6., 7., 8.],
[ 9., 10., 11., 12.],
[13., 14., 15., 16.]]]])In [60]:
pool2d = nn.MaxPool2d(3, padding=1, stride=2)
pool2d(X)Out [60]:
tensor([[[[ 5., 7.],
[13., 15.]],
[[ 6., 8.],
[14., 16.]]]])In [61]:
net = nn.Sequential(
nn.Conv2d(1,6,kernel_size=5,padding=2), #1*1*28*28 -> 1*6*28*28
nn.Sigmoid(),
nn.AvgPool2d(kernel_size=2, stride=2), #1*6*28*28 -> 1*6*14*14
nn.Conv2d(6, 16, kernel_size=5), nn.Sigmoid(), #1*6*14*14 -> 1*16*10*10
nn.AvgPool2d(kernel_size=2, stride=2), #1*16*10*10 -> 1*16*5*5
nn.Flatten(),
nn.Linear(16 * 5 * 5, 120), nn.Sigmoid(),
nn.Linear(120, 84), nn.Sigmoid(),
nn.Linear(84, 10)
)
X = torch.rand(size=(1,1,28,28),dtype=torch.float32)
for layer in net:
X=layer(X)
print(layer.__class__.__name__,'output shape: \t',X.shape)Conv2d output shape: torch.Size([1, 6, 28, 28]) Sigmoid output shape: torch.Size([1, 6, 28, 28]) AvgPool2d output shape: torch.Size([1, 6, 14, 14]) Conv2d output shape: torch.Size([1, 16, 10, 10]) Sigmoid output shape: torch.Size([1, 16, 10, 10]) AvgPool2d output shape: torch.Size([1, 16, 5, 5]) Flatten output shape: torch.Size([1, 400]) Linear output shape: torch.Size([1, 120]) Sigmoid output shape: torch.Size([1, 120]) Linear output shape: torch.Size([1, 84]) Sigmoid output shape: torch.Size([1, 84]) Linear output shape: torch.Size([1, 10])
In [62]:
import d2l.torch as d2l
batch_size = 256
train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size=batch_size)In [63]:
lr, num_epochs = 0.9, 10
#d2l.train_ch6(net, train_iter, test_iter, num_epochs, lr, d2l.try_gpu())In [64]:
class Inception(nn.Module):
def __init__(self,in_channels,c1,c2,c3,c4,**kwargs):
super(Inception,self).__init__(**kwargs)
self.p1_1 = nn.Conv2d(in_channels,c1,kernel_size=1)
self.p2_1 = nn.Conv2d(in_channels,c2[0],kernel_size=1)
self.p2_2 = nn.Conv2d(c2[0],c2[1],kernel_size=3,padding=1)
self.p3_1 = nn.Conv2d(in_channels,c3[0],kernel_size=1)
self.p3_2 = nn.Conv2d(c3[0],c3[1],kernel_size=5,padding=2)
self.p4_1 = nn.MaxPool2d(kernel_size=3, stride=1, padding=1)
self.p4_2 = nn.Conv2d(in_channels, c4, kernel_size=1)
def forward(self,x):
p1 = F.relu(self.p1_1(x))
p2 = F.relu(self.p2_2(F.relu(self.p2_1(x))))
p3 = F.relu(self.p3_2(F.relu(self.p3_1(x))))
p4 = F.relu(self.p4_2(self.p4_1(x)))
return torch.cat((p1,p2,p3,p4),dim=1)In [65]:
b1 = nn.Sequential(nn.Conv2d(1, 64, kernel_size=7, stride=2, padding=3),
nn.ReLU(),
nn.MaxPool2d(kernel_size=3, stride=2, padding=1))
b2 = nn.Sequential(nn.Conv2d(64, 64, kernel_size=1),
nn.ReLU(),
nn.Conv2d(64, 192, kernel_size=3, padding=1),
nn.ReLU(),
nn.MaxPool2d(kernel_size=3, stride=2, padding=1))
b3 = nn.Sequential(Inception(192, 64, (96, 128), (16, 32), 32),
Inception(256, 128, (128, 192), (32, 96), 64),
nn.MaxPool2d(kernel_size=3, stride=2, padding=1))
b4 = nn.Sequential(Inception(480, 192, (96, 208), (16, 48), 64),
Inception(512, 160, (112, 224), (24, 64), 64),
Inception(512, 128, (128, 256), (24, 64), 64),
Inception(512, 112, (144, 288), (32, 64), 64),
Inception(528, 256, (160, 320), (32, 128), 128),
nn.MaxPool2d(kernel_size=3, stride=2, padding=1))
b5 = nn.Sequential(Inception(832, 256, (160, 320), (32, 128), 128),
Inception(832, 384, (192, 384), (48, 128), 128),
nn.AdaptiveAvgPool2d((1,1)),
nn.Flatten())
net = nn.Sequential(b1, b2, b3, b4, b5, nn.Linear(1024, 10))
X = torch.rand(size=(1, 1, 96, 96))
for layer in net:
X = layer(X)
print(layer.__class__.__name__,'output shape:\t', X.shape)Sequential output shape: torch.Size([1, 64, 24, 24]) Sequential output shape: torch.Size([1, 192, 12, 12]) Sequential output shape: torch.Size([1, 480, 6, 6]) Sequential output shape: torch.Size([1, 832, 3, 3]) Sequential output shape: torch.Size([1, 1024]) Linear output shape: torch.Size([1, 10])
In [66]:
import torchinfo
torchinfo.summary(net,(1,1,96,96))Out [66]:
========================================================================================== Layer (type:depth-idx) Output Shape Param # ========================================================================================== Sequential [1, 10] -- ├─Sequential: 1-1 [1, 64, 24, 24] -- │ └─Conv2d: 2-1 [1, 64, 48, 48] 3,200 │ └─ReLU: 2-2 [1, 64, 48, 48] -- │ └─MaxPool2d: 2-3 [1, 64, 24, 24] -- ├─Sequential: 1-2 [1, 192, 12, 12] -- │ └─Conv2d: 2-4 [1, 64, 24, 24] 4,160 │ └─ReLU: 2-5 [1, 64, 24, 24] -- │ └─Conv2d: 2-6 [1, 192, 24, 24] 110,784 │ └─ReLU: 2-7 [1, 192, 24, 24] -- │ └─MaxPool2d: 2-8 [1, 192, 12, 12] -- ├─Sequential: 1-3 [1, 480, 6, 6] -- │ └─Inception: 2-9 [1, 256, 12, 12] -- │ │ └─Conv2d: 3-1 [1, 64, 12, 12] 12,352 │ │ └─Conv2d: 3-2 [1, 96, 12, 12] 18,528 │ │ └─Conv2d: 3-3 [1, 128, 12, 12] 110,720 │ │ └─Conv2d: 3-4 [1, 16, 12, 12] 3,088 │ │ └─Conv2d: 3-5 [1, 32, 12, 12] 12,832 │ │ └─MaxPool2d: 3-6 [1, 192, 12, 12] -- │ │ └─Conv2d: 3-7 [1, 32, 12, 12] 6,176 │ └─Inception: 2-10 [1, 480, 12, 12] -- │ │ └─Conv2d: 3-8 [1, 128, 12, 12] 32,896 │ │ └─Conv2d: 3-9 [1, 128, 12, 12] 32,896 │ │ └─Conv2d: 3-10 [1, 192, 12, 12] 221,376 │ │ └─Conv2d: 3-11 [1, 32, 12, 12] 8,224 │ │ └─Conv2d: 3-12 [1, 96, 12, 12] 76,896 │ │ └─MaxPool2d: 3-13 [1, 256, 12, 12] -- │ │ └─Conv2d: 3-14 [1, 64, 12, 12] 16,448 │ └─MaxPool2d: 2-11 [1, 480, 6, 6] -- ├─Sequential: 1-4 [1, 832, 3, 3] -- │ └─Inception: 2-12 [1, 512, 6, 6] -- │ │ └─Conv2d: 3-15 [1, 192, 6, 6] 92,352 │ │ └─Conv2d: 3-16 [1, 96, 6, 6] 46,176 │ │ └─Conv2d: 3-17 [1, 208, 6, 6] 179,920 │ │ └─Conv2d: 3-18 [1, 16, 6, 6] 7,696 │ │ └─Conv2d: 3-19 [1, 48, 6, 6] 19,248 │ │ └─MaxPool2d: 3-20 [1, 480, 6, 6] -- │ │ └─Conv2d: 3-21 [1, 64, 6, 6] 30,784 │ └─Inception: 2-13 [1, 512, 6, 6] -- │ │ └─Conv2d: 3-22 [1, 160, 6, 6] 82,080 │ │ └─Conv2d: 3-23 [1, 112, 6, 6] 57,456 │ │ └─Conv2d: 3-24 [1, 224, 6, 6] 226,016 │ │ └─Conv2d: 3-25 [1, 24, 6, 6] 12,312 │ │ └─Conv2d: 3-26 [1, 64, 6, 6] 38,464 │ │ └─MaxPool2d: 3-27 [1, 512, 6, 6] -- │ │ └─Conv2d: 3-28 [1, 64, 6, 6] 32,832 │ └─Inception: 2-14 [1, 512, 6, 6] -- │ │ └─Conv2d: 3-29 [1, 128, 6, 6] 65,664 │ │ └─Conv2d: 3-30 [1, 128, 6, 6] 65,664 │ │ └─Conv2d: 3-31 [1, 256, 6, 6] 295,168 │ │ └─Conv2d: 3-32 [1, 24, 6, 6] 12,312 │ │ └─Conv2d: 3-33 [1, 64, 6, 6] 38,464 │ │ └─MaxPool2d: 3-34 [1, 512, 6, 6] -- │ │ └─Conv2d: 3-35 [1, 64, 6, 6] 32,832 │ └─Inception: 2-15 [1, 528, 6, 6] -- │ │ └─Conv2d: 3-36 [1, 112, 6, 6] 57,456 │ │ └─Conv2d: 3-37 [1, 144, 6, 6] 73,872 │ │ └─Conv2d: 3-38 [1, 288, 6, 6] 373,536 │ │ └─Conv2d: 3-39 [1, 32, 6, 6] 16,416 │ │ └─Conv2d: 3-40 [1, 64, 6, 6] 51,264 │ │ └─MaxPool2d: 3-41 [1, 512, 6, 6] -- │ │ └─Conv2d: 3-42 [1, 64, 6, 6] 32,832 │ └─Inception: 2-16 [1, 832, 6, 6] -- │ │ └─Conv2d: 3-43 [1, 256, 6, 6] 135,424 │ │ └─Conv2d: 3-44 [1, 160, 6, 6] 84,640 │ │ └─Conv2d: 3-45 [1, 320, 6, 6] 461,120 │ │ └─Conv2d: 3-46 [1, 32, 6, 6] 16,928 │ │ └─Conv2d: 3-47 [1, 128, 6, 6] 102,528 │ │ └─MaxPool2d: 3-48 [1, 528, 6, 6] -- │ │ └─Conv2d: 3-49 [1, 128, 6, 6] 67,712 │ └─MaxPool2d: 2-17 [1, 832, 3, 3] -- ├─Sequential: 1-5 [1, 1024] -- │ └─Inception: 2-18 [1, 832, 3, 3] -- │ │ └─Conv2d: 3-50 [1, 256, 3, 3] 213,248 │ │ └─Conv2d: 3-51 [1, 160, 3, 3] 133,280 │ │ └─Conv2d: 3-52 [1, 320, 3, 3] 461,120 │ │ └─Conv2d: 3-53 [1, 32, 3, 3] 26,656 │ │ └─Conv2d: 3-54 [1, 128, 3, 3] 102,528 │ │ └─MaxPool2d: 3-55 [1, 832, 3, 3] -- │ │ └─Conv2d: 3-56 [1, 128, 3, 3] 106,624 │ └─Inception: 2-19 [1, 1024, 3, 3] -- │ │ └─Conv2d: 3-57 [1, 384, 3, 3] 319,872 │ │ └─Conv2d: 3-58 [1, 192, 3, 3] 159,936 │ │ └─Conv2d: 3-59 [1, 384, 3, 3] 663,936 │ │ └─Conv2d: 3-60 [1, 48, 3, 3] 39,984 │ │ └─Conv2d: 3-61 [1, 128, 3, 3] 153,728 │ │ └─MaxPool2d: 3-62 [1, 832, 3, 3] -- │ │ └─Conv2d: 3-63 [1, 128, 3, 3] 106,624 │ └─AdaptiveAvgPool2d: 2-20 [1, 1024, 1, 1] -- │ └─Flatten: 2-21 [1, 1024] -- ├─Linear: 1-6 [1, 10] 10,250 ========================================================================================== Total params: 5,977,530 Trainable params: 5,977,530 Non-trainable params: 0 Total mult-adds (Units.MEGABYTES): 276.66 ========================================================================================== Input size (MB): 0.04 Forward/backward pass size (MB): 4.74 Params size (MB): 23.91 Estimated Total Size (MB): 28.69 ==========================================================================================
In [67]:
lr, num_epochs, batch_size = 0.1, 10, 128
train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size, resize=96)
#d2l.train_ch6(net, train_iter, test_iter, num_epochs, lr, d2l.try_gpu())In [68]:
class Residual(nn.Module):
def __init__(self,input_channels,num_channels,use_1x1conv=False,strides=1):
super().__init__()
self.conv1 = nn.Conv2d(input_channels,num_channels,kernel_size=3,padding=1,stride=strides)
self.conv2 = nn.Conv2d(num_channels,num_channels,kernel_size=3,padding=1)
if use_1x1conv:
self.conv3 = nn.Conv2d(input_channels,num_channels,kernel_size=1,stride=strides)
else:
self.conv3= None
self.bn1=nn.BatchNorm2d(num_channels)
self.bn2=nn.BatchNorm2d(num_channels)
def forward(self,X):
Y=F.relu(self.bn1(self.conv1(X)))
Y=self.bn2(self.conv2(Y))
if self.conv3:
X = self.conv3(X)
Y+=X
return F.relu(Y)
In [69]:
blk = Residual(3,3)
X = torch.rand(4, 3, 6, 6)In [70]:
blk = Residual(3,6, use_1x1conv=True, strides=2)
blk(X).shapeOut [70]:
torch.Size([4, 6, 3, 3])
In [71]:
b1 = nn.Sequential(nn.Conv2d(1, 64, kernel_size=7, stride=2, padding=3),
nn.BatchNorm2d(64), nn.ReLU(),
nn.MaxPool2d(kernel_size=3, stride=2, padding=1))In [72]:
def resnet_block(input_channels, num_channels, num_residuals,
first_block=False):
blk = []
for i in range(num_residuals):
if i == 0 and not first_block:
blk.append(Residual(input_channels, num_channels,
use_1x1conv=True, strides=2))
else:
blk.append(Residual(num_channels, num_channels))
return blkIn [73]:
b2 = nn.Sequential(*resnet_block(64, 64, 2, first_block=True))
b3 = nn.Sequential(*resnet_block(64, 128, 2))
b4 = nn.Sequential(*resnet_block(128, 256, 2))
b5 = nn.Sequential(*resnet_block(256, 512, 2))In [74]:
net = nn.Sequential(b1, b2, b3, b4, b5,
nn.AdaptiveAvgPool2d((1,1)),
nn.Flatten(), nn.Linear(512, 10))In [75]:
X = torch.rand(size=(1, 1, 224, 224))
for layer in net:
X = layer(X)
print(layer.__class__.__name__,'output shape:\t', X.shape)Sequential output shape: torch.Size([1, 64, 56, 56]) Sequential output shape: torch.Size([1, 64, 56, 56]) Sequential output shape: torch.Size([1, 128, 28, 28]) Sequential output shape: torch.Size([1, 256, 14, 14]) Sequential output shape: torch.Size([1, 512, 7, 7]) AdaptiveAvgPool2d output shape: torch.Size([1, 512, 1, 1]) Flatten output shape: torch.Size([1, 512]) Linear output shape: torch.Size([1, 10])
In [76]:
lr, num_epochs, batch_size = 0.05, 10, 256
train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size, resize=96)
#d2l.train_ch6(net, train_iter, test_iter, num_epochs, lr, d2l.try_gpu())In [77]:
import torch
import d2l.torch as d2l
import numpy
import torch.nn as nn
import torch.nn.functional as F
print(torch.version.__version__)2.10.0+cu128
In [78]:
A=torch.Tensor([[1,2,0,0],[0,2,0,0],[0,0,2,1],[0,0,0,3]])
C=torch.Tensor([[1,0,0,0],[0,1,0,0],[0,0,-2,3],[0,0,0,-3]])In [79]:
B=torch.Tensor([[2,0,0,0],[-2,1,0,0],[0,0,-3,0],[0,0,0,-3]])In [80]:
torch.mm(A,C)Out [80]:
tensor([[ 1., 2., 0., 0.],
[ 0., 2., 0., 0.],
[ 0., 0., -4., 3.],
[ 0., 0., 0., -9.]])In [81]:
torch.det(torch.mm(torch.mm(A,C),B))Out [81]:
tensor(1296.)
In [82]:
1296**5
Out [82]:
3656158440062976
In [83]:
torch.mm(C,B)Out [83]:
tensor([[ 2., 0., 0., 0.],
[-2., 1., 0., 0.],
[ 0., 0., 6., -9.],
[ 0., 0., 0., 9.]])In [84]:
T = 1000 # 总共产生1000个点
time = torch.arange(1, T + 1, dtype=torch.float32)
x = torch.sin(0.01 * time) + torch.normal(0, 0.2, (T,))
d2l.plot(time, [x], 'time', 'x', xlim=[1, 1000], figsize=(6, 3))
In [85]:
tau = 4
features = torch.zeros((T - tau, tau))
for i in range(tau):
features[:, i] = x[i: T - tau + i]
labels = x[tau:].reshape((-1, 1))
x,features,labels
Out [85]:
(tensor([-0.0948, 0.2143, -0.2523, -0.1235, -0.1826, 0.1189, -0.1963, 0.2347,
0.1456, -0.1118, 0.3787, 0.3861, 0.2881, 0.1958, 0.0402, 0.0816,
0.4793, 0.0351, 0.2378, 0.1459, 0.1108, 0.2544, -0.0127, 0.0733,
0.3156, 0.0257, 0.3207, 0.3259, 0.3693, 0.0584, 0.1730, 0.3100,
0.2328, 0.0525, 0.4465, 0.1293, 0.4330, 0.3193, 0.4704, 0.5238,
0.5323, 0.4887, 0.0831, 0.5924, 0.6972, 0.3490, 0.7476, 0.6039,
0.9995, 0.1455, 0.1417, 0.5968, 0.6673, 0.3425, 0.7685, 0.4904,
0.2203, 0.2109, 0.4600, 0.5055, 0.3558, 0.7020, 0.7435, 0.4713,
0.4318, 0.5861, 0.3592, 0.7750, 0.6640, 0.7908, 0.2776, 0.5868,
0.6283, 0.3461, 0.6308, 0.7547, 0.5564, 0.7181, 0.7852, 0.7823,
0.7238, 0.9294, 0.9023, 0.8100, 0.5561, 0.7124, 1.1566, 0.7628,
0.9630, 0.4425, 1.0628, 0.7014, 0.4439, 0.7286, 0.8099, 0.5786,
1.0638, 0.9519, 0.8388, 1.2088, 0.9172, 0.7014, 0.5667, 0.6040,
0.5549, 0.7959, 0.9167, 0.9074, 0.6108, 0.8999, 0.9197, 0.8539,
0.6566, 0.9941, 0.6902, 0.8782, 1.4898, 0.9888, 1.1911, 0.5683,
0.8868, 0.7122, 0.8960, 1.1454, 1.2660, 1.0001, 0.6582, 0.9706,
1.0110, 1.0355, 0.9761, 0.9439, 1.0824, 1.4095, 1.2544, 0.6541,
1.0486, 1.0638, 0.9000, 0.9835, 1.2558, 1.1702, 0.8466, 0.7696,
1.2446, 1.1460, 0.9258, 0.8150, 1.1086, 0.9475, 0.9675, 0.7330,
1.1263, 1.1718, 0.9413, 1.0272, 0.7733, 0.9831, 0.8759, 0.7970,
0.6360, 1.1815, 0.9689, 0.6976, 0.9265, 0.8338, 0.7960, 0.7705,
1.2601, 1.2775, 0.7706, 1.0216, 1.1916, 0.8603, 0.9864, 1.0777,
0.8930, 1.0063, 0.8376, 0.9923, 0.8081, 0.8020, 1.1461, 1.1018,
0.8931, 1.0005, 0.8635, 0.7197, 1.2577, 1.0584, 1.4032, 0.8911,
1.1415, 0.8241, 0.7946, 1.0221, 0.8792, 0.7211, 1.1821, 0.8079,
0.8926, 1.0765, 0.9949, 0.9159, 0.7329, 0.9950, 0.7491, 0.8750,
1.1863, 1.0095, 0.8046, 0.6274, 0.8936, 0.7595, 0.8423, 0.8655,
0.6918, 0.7347, 1.1179, 0.5931, 0.8745, 0.4858, 0.9338, 1.1382,
0.6084, 0.9479, 0.8726, 0.7202, 0.9596, 0.4386, 1.2525, 0.5120,
0.7222, 0.6566, 0.8965, 0.7545, 1.1104, 0.6634, 0.5654, 1.0095,
0.6558, 0.7260, 0.8515, 0.3430, 0.7703, 0.3753, 0.4490, 0.4373,
0.8283, 0.5455, 0.7584, 0.8197, 0.4781, 0.3350, 0.6714, 0.3969,
0.7131, 0.5609, 0.4327, 0.4293, 0.3552, 0.5445, 0.5609, 0.4110,
0.8525, 0.3402, 0.4064, 0.5172, 0.5845, 0.6185, 0.4719, 0.9092,
0.6964, 0.7267, 0.6934, 0.4337, 0.2031, 0.0898, 0.4377, 0.4203,
0.2855, 0.4673, 0.6029, 0.4368, 0.1521, -0.0606, 0.2532, 0.4365,
0.2989, 0.0743, 0.2734, 0.1060, 0.5543, -0.1211, 0.0968, 0.4911,
0.5107, 0.4583, 0.2777, -0.0513, 0.1437, 0.0548, 0.0933, 0.1172,
0.0718, 0.4027, 0.1805, 0.0869, -0.3066, 0.5615, -0.2721, -0.2765,
0.0850, -0.1473, -0.1622, -0.1335, 0.1328, 0.0703, -0.6712, -0.1121,
-0.1208, 0.0092, -0.0805, -0.2017, 0.2339, -0.3533, -0.4598, 0.0620,
-0.5254, 0.0197, -0.0593, -0.1914, -0.4259, -0.0115, -0.5406, 0.0137,
-0.4240, -0.2822, -0.0796, -0.3495, -0.4475, 0.2453, -0.3729, -0.4086,
-0.2618, -0.4539, -0.6140, -0.2483, -0.4165, -0.3736, -0.0737, -0.0212,
-0.3644, -0.0472, -0.4087, -0.6794, -0.5921, -0.5632, -0.5971, -0.1935,
-0.9543, -0.7976, -0.3485, -0.9538, -0.6171, -0.7755, -0.4651, -0.8194,
-0.3005, -0.5191, -0.5902, -0.2464, -0.6908, -0.5054, -0.5528, -1.1089,
-0.7206, -0.8067, -0.6780, -0.2981, -0.6683, -0.4324, -0.8497, -0.5928,
-0.7203, -0.3751, -0.7423, -0.4109, -0.7345, -0.6653, -0.5752, -0.5198,
-0.7046, -1.1754, -0.9447, -0.7304, -0.6510, -0.5954, -0.7592, -0.5285,
-0.4249, -0.7993, -1.3758, -0.6218, -1.0691, -0.5775, -0.8174, -0.7021,
-0.7784, -0.7553, -1.2137, -0.7302, -0.7253, -0.6819, -1.3077, -1.3472,
-0.7104, -0.8387, -0.5973, -0.8619, -1.1138, -1.1314, -0.9765, -1.2121,
-0.8168, -0.7763, -1.2988, -0.9282, -1.1715, -0.7216, -0.7182, -0.2972,
-0.7471, -1.0089, -1.1431, -1.0396, -1.0381, -0.5979, -0.7363, -0.7808,
-0.9106, -1.1468, -1.1357, -0.6406, -0.9603, -1.2653, -1.5958, -1.0592,
-0.9698, -0.8252, -1.2515, -1.0474, -1.1103, -1.0035, -0.6669, -0.9120,
-0.9146, -1.1079, -0.8379, -0.9123, -0.5831, -1.6515, -0.9385, -1.0699,
-1.1498, -0.7861, -0.8942, -1.0452, -1.0064, -0.9116, -1.1150, -0.7801,
-1.0283, -1.0296, -1.0927, -0.7945, -1.0705, -1.3215, -1.2510, -0.9158,
-0.9377, -0.7314, -0.9773, -1.1910, -1.0539, -1.1439, -1.0784, -0.8543,
-1.1323, -1.3193, -0.8014, -0.7318, -0.5805, -0.8239, -1.1228, -1.0473,
-0.8206, -0.6544, -1.2654, -1.0757, -0.5389, -0.9908, -0.7894, -0.7463,
-1.0391, -0.8023, -0.8568, -1.2414, -0.9595, -1.1151, -0.9689, -1.1145,
-0.6853, -0.7547, -1.1000, -0.9054, -1.2262, -1.1359, -1.0174, -0.3782,
-0.8056, -1.1828, -0.8426, -0.9958, -0.9495, -1.2745, -0.7039, -0.5893,
-0.5648, -1.0538, -0.6724, -0.6340, -0.5070, -1.0956, -1.0957, -0.6823,
-0.5258, -0.5777, -0.9268, -0.5280, -0.5989, -0.8364, -0.7439, -0.7619,
-1.0159, -1.0627, -0.9416, -0.6270, -0.4307, -0.8575, -1.0748, -0.5529,
-0.9339, -0.7416, -0.6674, -0.3178, -0.6815, -0.7499, -0.6359, -0.8157,
-0.5582, -0.5083, -0.4527, -0.8350, -0.6317, -0.4338, -0.4875, -0.4046,
-0.3166, -0.3413, -0.4722, -0.7010, -1.2025, -0.2133, -0.3133, -0.4160,
-0.6681, -0.8990, -0.5464, -0.4518, -0.4402, -0.5246, -0.4561, -0.6747,
-0.1833, -0.4466, -0.4671, -0.5509, -0.6235, -0.2100, -0.3368, -0.3083,
-0.5129, -0.2880, -0.4075, -0.2784, 0.0631, -0.4355, -0.4237, -0.2578,
-0.1380, -0.5085, 0.1004, -0.1426, -0.2537, -0.1756, -0.2135, -0.1898,
-0.2947, -0.3934, -0.3412, -0.3343, -0.1450, -0.3178, -0.2156, -0.3232,
-0.3691, -0.2711, 0.1086, -0.2257, -0.0752, -0.0339, -0.0636, 0.0626,
-0.1460, 0.0792, 0.1529, 0.4743, 0.0343, -0.0158, -0.1255, -0.4698,
-0.0489, 0.2622, 0.0619, -0.2243, -0.1318, 0.0214, 0.2690, 0.0497,
0.3451, -0.1116, 0.0173, 0.0708, 0.4135, 0.3188, 0.4808, -0.0340,
0.4786, 0.4896, 0.1077, 0.3500, 0.1309, 0.1398, 0.1943, 0.1651,
0.3227, 0.5541, 0.2688, 0.1892, 0.2509, 0.2078, -0.0140, 0.2443,
0.3204, 0.5485, 0.4234, 0.3135, 0.4633, 0.0029, 0.2174, 0.6879,
0.5089, 0.2479, 0.8608, 0.4307, 0.6205, 0.3482, 0.6469, 0.4475,
0.6595, 0.3450, 0.3781, 0.4451, 0.1883, 0.6707, 0.8667, 0.5218,
0.4004, 0.5271, 0.6446, 0.7222, 0.5722, 0.7676, 0.6824, 0.1981,
0.8089, 0.6296, 0.6748, 0.7515, 0.5103, 0.9052, 0.8405, 0.9092,
0.6918, 0.6477, 0.5402, 0.6477, 0.4210, 0.6973, 0.6019, 0.5364,
0.8134, 0.5607, 0.7096, 0.5894, 0.3866, 1.0600, 0.7347, 0.8129,
1.2088, 0.8825, 0.7179, 1.0115, 0.7013, 1.0128, 0.9747, 1.2759,
0.7655, 1.0094, 0.7805, 0.6091, 1.2033, 0.9678, 0.8219, 0.8157,
0.9188, 0.7436, 0.8910, 0.7291, 0.9559, 0.9389, 1.2030, 1.0495,
1.1811, 0.8884, 0.8390, 0.9894, 0.9238, 0.7628, 0.5421, 1.5147,
0.6971, 0.6740, 0.8342, 0.6554, 0.7455, 0.6916, 1.2706, 1.1277,
0.9248, 0.9976, 1.2404, 0.6919, 1.3449, 1.1243, 1.0492, 0.9266,
1.1194, 1.0304, 1.1323, 1.2372, 0.8300, 1.1916, 1.0923, 0.8313,
0.8572, 1.1128, 1.0047, 1.1544, 0.9745, 1.0503, 0.9171, 0.8073,
1.2056, 1.0976, 0.9910, 1.1834, 1.1389, 0.9142, 0.9367, 1.0121,
0.7704, 1.0558, 0.7306, 0.8117, 0.7061, 1.2315, 0.9015, 0.9339,
0.5016, 0.9227, 1.2568, 0.9444, 1.1198, 0.9431, 1.0997, 1.3078,
0.8336, 1.2692, 0.8424, 0.8702, 1.4820, 1.3248, 0.9324, 0.6538,
1.2011, 1.0170, 0.7863, 1.0178, 0.6519, 0.5970, 0.9052, 0.6846,
0.7737, 0.9104, 0.8439, 1.0066, 1.0787, 0.9661, 0.9923, 0.7922,
0.8316, 0.9553, 0.9952, 0.8680, 1.1226, 0.8213, 0.9151, 0.7748,
0.9953, 0.7773, 0.7916, 0.7321, 0.9130, 1.1433, 0.7060, 0.8066,
0.8709, 0.7426, 0.8718, 1.0973, 0.7097, 0.9438, 0.8164, 0.8013,
0.6236, 0.7180, 0.9188, 0.8016, 0.9741, 0.6271, 0.5747, 0.8007,
0.7754, 0.4877, 0.4746, 0.8654, 0.4743, 0.9015, 0.8082, 0.5449,
0.9299, 0.2003, 0.5466, 0.4355, 0.7900, 0.4343, 0.7224, 0.8585,
0.5714, 0.5306, 0.6594, 0.0640, 0.3203, 0.5463, 0.5048, 0.1935,
0.2883, 0.6778, 0.5014, 0.5235, 0.5718, 0.4587, 0.2808, 0.4073,
0.8632, 0.8862, 0.5757, 0.3372, 0.2566, 0.7858, 0.3713, 0.1589,
0.3243, 0.4270, 0.0565, 0.2885, 0.3257, 0.2196, 0.3159, 0.2361,
0.1087, 0.2224, 0.2633, 0.5037, 0.1980, 0.1530, 0.2780, -0.1399,
0.5331, 0.3530, 0.3342, 0.2098, -0.0165, 0.1318, 0.4510, -0.1959,
0.0966, 0.0789, 0.3381, -0.1917, 0.1518, 0.3640, 0.0956, 0.2535,
-0.3988, -0.3479, 0.3864, -0.2639, -0.2368, 0.0258, 0.2441, 0.0687,
0.0457, 0.2286, -0.0947, -0.1189, 0.1360, -0.0990, -0.2447, 0.2135,
-0.1830, -0.4583, -0.1795, -0.1361, -0.0553, -0.2864, -0.2307, -0.4651,
-0.1889, -0.3185, -0.5318, -0.3012, 0.0062, 0.1046, -0.2321, -0.2945,
-0.0242, -0.0586, -0.2307, -0.2479, -0.0382, -0.1509, -0.5055, -0.3759,
0.2139, -0.2129, -0.3605, -0.5222, -0.6530, -0.6716, -0.4330, -0.2577,
-0.2672, -0.1297, -0.9203, -0.5832, -0.2640, -0.4996, -0.2625, -0.4407,
-0.8864, -0.2508, -0.4827, -0.3131, -0.2570, -0.7116, -0.5357, -0.7074]),
tensor([[-0.0948, 0.2143, -0.2523, -0.1235],
[ 0.2143, -0.2523, -0.1235, -0.1826],
[-0.2523, -0.1235, -0.1826, 0.1189],
...,
[-0.2508, -0.4827, -0.3131, -0.2570],
[-0.4827, -0.3131, -0.2570, -0.7116],
[-0.3131, -0.2570, -0.7116, -0.5357]]),
tensor([[-0.1826],
[ 0.1189],
[-0.1963],
[ 0.2347],
[ 0.1456],
[-0.1118],
[ 0.3787],
[ 0.3861],
[ 0.2881],
[ 0.1958],
[ 0.0402],
[ 0.0816],
[ 0.4793],
[ 0.0351],
[ 0.2378],
[ 0.1459],
[ 0.1108],
[ 0.2544],
[-0.0127],
[ 0.0733],
[ 0.3156],
[ 0.0257],
[ 0.3207],
[ 0.3259],
[ 0.3693],
[ 0.0584],
[ 0.1730],
[ 0.3100],
[ 0.2328],
[ 0.0525],
[ 0.4465],
[ 0.1293],
[ 0.4330],
[ 0.3193],
[ 0.4704],
[ 0.5238],
[ 0.5323],
[ 0.4887],
[ 0.0831],
[ 0.5924],
[ 0.6972],
[ 0.3490],
[ 0.7476],
[ 0.6039],
[ 0.9995],
[ 0.1455],
[ 0.1417],
[ 0.5968],
[ 0.6673],
[ 0.3425],
[ 0.7685],
[ 0.4904],
[ 0.2203],
[ 0.2109],
[ 0.4600],
[ 0.5055],
[ 0.3558],
[ 0.7020],
[ 0.7435],
[ 0.4713],
[ 0.4318],
[ 0.5861],
[ 0.3592],
[ 0.7750],
[ 0.6640],
[ 0.7908],
[ 0.2776],
[ 0.5868],
[ 0.6283],
[ 0.3461],
[ 0.6308],
[ 0.7547],
[ 0.5564],
[ 0.7181],
[ 0.7852],
[ 0.7823],
[ 0.7238],
[ 0.9294],
[ 0.9023],
[ 0.8100],
[ 0.5561],
[ 0.7124],
[ 1.1566],
[ 0.7628],
[ 0.9630],
[ 0.4425],
[ 1.0628],
[ 0.7014],
[ 0.4439],
[ 0.7286],
[ 0.8099],
[ 0.5786],
[ 1.0638],
[ 0.9519],
[ 0.8388],
[ 1.2088],
[ 0.9172],
[ 0.7014],
[ 0.5667],
[ 0.6040],
[ 0.5549],
[ 0.7959],
[ 0.9167],
[ 0.9074],
[ 0.6108],
[ 0.8999],
[ 0.9197],
[ 0.8539],
[ 0.6566],
[ 0.9941],
[ 0.6902],
[ 0.8782],
[ 1.4898],
[ 0.9888],
[ 1.1911],
[ 0.5683],
[ 0.8868],
[ 0.7122],
[ 0.8960],
[ 1.1454],
[ 1.2660],
[ 1.0001],
[ 0.6582],
[ 0.9706],
[ 1.0110],
[ 1.0355],
[ 0.9761],
[ 0.9439],
[ 1.0824],
[ 1.4095],
[ 1.2544],
[ 0.6541],
[ 1.0486],
[ 1.0638],
[ 0.9000],
[ 0.9835],
[ 1.2558],
[ 1.1702],
[ 0.8466],
[ 0.7696],
[ 1.2446],
[ 1.1460],
[ 0.9258],
[ 0.8150],
[ 1.1086],
[ 0.9475],
[ 0.9675],
[ 0.7330],
[ 1.1263],
[ 1.1718],
[ 0.9413],
[ 1.0272],
[ 0.7733],
[ 0.9831],
[ 0.8759],
[ 0.7970],
[ 0.6360],
[ 1.1815],
[ 0.9689],
[ 0.6976],
[ 0.9265],
[ 0.8338],
[ 0.7960],
[ 0.7705],
[ 1.2601],
[ 1.2775],
[ 0.7706],
[ 1.0216],
[ 1.1916],
[ 0.8603],
[ 0.9864],
[ 1.0777],
[ 0.8930],
[ 1.0063],
[ 0.8376],
[ 0.9923],
[ 0.8081],
[ 0.8020],
[ 1.1461],
[ 1.1018],
[ 0.8931],
[ 1.0005],
[ 0.8635],
[ 0.7197],
[ 1.2577],
[ 1.0584],
[ 1.4032],
[ 0.8911],
[ 1.1415],
[ 0.8241],
[ 0.7946],
[ 1.0221],
[ 0.8792],
[ 0.7211],
[ 1.1821],
[ 0.8079],
[ 0.8926],
[ 1.0765],
[ 0.9949],
[ 0.9159],
[ 0.7329],
[ 0.9950],
[ 0.7491],
[ 0.8750],
[ 1.1863],
[ 1.0095],
[ 0.8046],
[ 0.6274],
[ 0.8936],
[ 0.7595],
[ 0.8423],
[ 0.8655],
[ 0.6918],
[ 0.7347],
[ 1.1179],
[ 0.5931],
[ 0.8745],
[ 0.4858],
[ 0.9338],
[ 1.1382],
[ 0.6084],
[ 0.9479],
[ 0.8726],
[ 0.7202],
[ 0.9596],
[ 0.4386],
[ 1.2525],
[ 0.5120],
[ 0.7222],
[ 0.6566],
[ 0.8965],
[ 0.7545],
[ 1.1104],
[ 0.6634],
[ 0.5654],
[ 1.0095],
[ 0.6558],
[ 0.7260],
[ 0.8515],
[ 0.3430],
[ 0.7703],
[ 0.3753],
[ 0.4490],
[ 0.4373],
[ 0.8283],
[ 0.5455],
[ 0.7584],
[ 0.8197],
[ 0.4781],
[ 0.3350],
[ 0.6714],
[ 0.3969],
[ 0.7131],
[ 0.5609],
[ 0.4327],
[ 0.4293],
[ 0.3552],
[ 0.5445],
[ 0.5609],
[ 0.4110],
[ 0.8525],
[ 0.3402],
[ 0.4064],
[ 0.5172],
[ 0.5845],
[ 0.6185],
[ 0.4719],
[ 0.9092],
[ 0.6964],
[ 0.7267],
[ 0.6934],
[ 0.4337],
[ 0.2031],
[ 0.0898],
[ 0.4377],
[ 0.4203],
[ 0.2855],
[ 0.4673],
[ 0.6029],
[ 0.4368],
[ 0.1521],
[-0.0606],
[ 0.2532],
[ 0.4365],
[ 0.2989],
[ 0.0743],
[ 0.2734],
[ 0.1060],
[ 0.5543],
[-0.1211],
[ 0.0968],
[ 0.4911],
[ 0.5107],
[ 0.4583],
[ 0.2777],
[-0.0513],
[ 0.1437],
[ 0.0548],
[ 0.0933],
[ 0.1172],
[ 0.0718],
[ 0.4027],
[ 0.1805],
[ 0.0869],
[-0.3066],
[ 0.5615],
[-0.2721],
[-0.2765],
[ 0.0850],
[-0.1473],
[-0.1622],
[-0.1335],
[ 0.1328],
[ 0.0703],
[-0.6712],
[-0.1121],
[-0.1208],
[ 0.0092],
[-0.0805],
[-0.2017],
[ 0.2339],
[-0.3533],
[-0.4598],
[ 0.0620],
[-0.5254],
[ 0.0197],
[-0.0593],
[-0.1914],
[-0.4259],
[-0.0115],
[-0.5406],
[ 0.0137],
[-0.4240],
[-0.2822],
[-0.0796],
[-0.3495],
[-0.4475],
[ 0.2453],
[-0.3729],
[-0.4086],
[-0.2618],
[-0.4539],
[-0.6140],
[-0.2483],
[-0.4165],
[-0.3736],
[-0.0737],
[-0.0212],
[-0.3644],
[-0.0472],
[-0.4087],
[-0.6794],
[-0.5921],
[-0.5632],
[-0.5971],
[-0.1935],
[-0.9543],
[-0.7976],
[-0.3485],
[-0.9538],
[-0.6171],
[-0.7755],
[-0.4651],
[-0.8194],
[-0.3005],
[-0.5191],
[-0.5902],
[-0.2464],
[-0.6908],
[-0.5054],
[-0.5528],
[-1.1089],
[-0.7206],
[-0.8067],
[-0.6780],
[-0.2981],
[-0.6683],
[-0.4324],
[-0.8497],
[-0.5928],
[-0.7203],
[-0.3751],
[-0.7423],
[-0.4109],
[-0.7345],
[-0.6653],
[-0.5752],
[-0.5198],
[-0.7046],
[-1.1754],
[-0.9447],
[-0.7304],
[-0.6510],
[-0.5954],
[-0.7592],
[-0.5285],
[-0.4249],
[-0.7993],
[-1.3758],
[-0.6218],
[-1.0691],
[-0.5775],
[-0.8174],
[-0.7021],
[-0.7784],
[-0.7553],
[-1.2137],
[-0.7302],
[-0.7253],
[-0.6819],
[-1.3077],
[-1.3472],
[-0.7104],
[-0.8387],
[-0.5973],
[-0.8619],
[-1.1138],
[-1.1314],
[-0.9765],
[-1.2121],
[-0.8168],
[-0.7763],
[-1.2988],
[-0.9282],
[-1.1715],
[-0.7216],
[-0.7182],
[-0.2972],
[-0.7471],
[-1.0089],
[-1.1431],
[-1.0396],
[-1.0381],
[-0.5979],
[-0.7363],
[-0.7808],
[-0.9106],
[-1.1468],
[-1.1357],
[-0.6406],
[-0.9603],
[-1.2653],
[-1.5958],
[-1.0592],
[-0.9698],
[-0.8252],
[-1.2515],
[-1.0474],
[-1.1103],
[-1.0035],
[-0.6669],
[-0.9120],
[-0.9146],
[-1.1079],
[-0.8379],
[-0.9123],
[-0.5831],
[-1.6515],
[-0.9385],
[-1.0699],
[-1.1498],
[-0.7861],
[-0.8942],
[-1.0452],
[-1.0064],
[-0.9116],
[-1.1150],
[-0.7801],
[-1.0283],
[-1.0296],
[-1.0927],
[-0.7945],
[-1.0705],
[-1.3215],
[-1.2510],
[-0.9158],
[-0.9377],
[-0.7314],
[-0.9773],
[-1.1910],
[-1.0539],
[-1.1439],
[-1.0784],
[-0.8543],
[-1.1323],
[-1.3193],
[-0.8014],
[-0.7318],
[-0.5805],
[-0.8239],
[-1.1228],
[-1.0473],
[-0.8206],
[-0.6544],
[-1.2654],
[-1.0757],
[-0.5389],
[-0.9908],
[-0.7894],
[-0.7463],
[-1.0391],
[-0.8023],
[-0.8568],
[-1.2414],
[-0.9595],
[-1.1151],
[-0.9689],
[-1.1145],
[-0.6853],
[-0.7547],
[-1.1000],
[-0.9054],
[-1.2262],
[-1.1359],
[-1.0174],
[-0.3782],
[-0.8056],
[-1.1828],
[-0.8426],
[-0.9958],
[-0.9495],
[-1.2745],
[-0.7039],
[-0.5893],
[-0.5648],
[-1.0538],
[-0.6724],
[-0.6340],
[-0.5070],
[-1.0956],
[-1.0957],
[-0.6823],
[-0.5258],
[-0.5777],
[-0.9268],
[-0.5280],
[-0.5989],
[-0.8364],
[-0.7439],
[-0.7619],
[-1.0159],
[-1.0627],
[-0.9416],
[-0.6270],
[-0.4307],
[-0.8575],
[-1.0748],
[-0.5529],
[-0.9339],
[-0.7416],
[-0.6674],
[-0.3178],
[-0.6815],
[-0.7499],
[-0.6359],
[-0.8157],
[-0.5582],
[-0.5083],
[-0.4527],
[-0.8350],
[-0.6317],
[-0.4338],
[-0.4875],
[-0.4046],
[-0.3166],
[-0.3413],
[-0.4722],
[-0.7010],
[-1.2025],
[-0.2133],
[-0.3133],
[-0.4160],
[-0.6681],
[-0.8990],
[-0.5464],
[-0.4518],
[-0.4402],
[-0.5246],
[-0.4561],
[-0.6747],
[-0.1833],
[-0.4466],
[-0.4671],
[-0.5509],
[-0.6235],
[-0.2100],
[-0.3368],
[-0.3083],
[-0.5129],
[-0.2880],
[-0.4075],
[-0.2784],
[ 0.0631],
[-0.4355],
[-0.4237],
[-0.2578],
[-0.1380],
[-0.5085],
[ 0.1004],
[-0.1426],
[-0.2537],
[-0.1756],
[-0.2135],
[-0.1898],
[-0.2947],
[-0.3934],
[-0.3412],
[-0.3343],
[-0.1450],
[-0.3178],
[-0.2156],
[-0.3232],
[-0.3691],
[-0.2711],
[ 0.1086],
[-0.2257],
[-0.0752],
[-0.0339],
[-0.0636],
[ 0.0626],
[-0.1460],
[ 0.0792],
[ 0.1529],
[ 0.4743],
[ 0.0343],
[-0.0158],
[-0.1255],
[-0.4698],
[-0.0489],
[ 0.2622],
[ 0.0619],
[-0.2243],
[-0.1318],
[ 0.0214],
[ 0.2690],
[ 0.0497],
[ 0.3451],
[-0.1116],
[ 0.0173],
[ 0.0708],
[ 0.4135],
[ 0.3188],
[ 0.4808],
[-0.0340],
[ 0.4786],
[ 0.4896],
[ 0.1077],
[ 0.3500],
[ 0.1309],
[ 0.1398],
[ 0.1943],
[ 0.1651],
[ 0.3227],
[ 0.5541],
[ 0.2688],
[ 0.1892],
[ 0.2509],
[ 0.2078],
[-0.0140],
[ 0.2443],
[ 0.3204],
[ 0.5485],
[ 0.4234],
[ 0.3135],
[ 0.4633],
[ 0.0029],
[ 0.2174],
[ 0.6879],
[ 0.5089],
[ 0.2479],
[ 0.8608],
[ 0.4307],
[ 0.6205],
[ 0.3482],
[ 0.6469],
[ 0.4475],
[ 0.6595],
[ 0.3450],
[ 0.3781],
[ 0.4451],
[ 0.1883],
[ 0.6707],
[ 0.8667],
[ 0.5218],
[ 0.4004],
[ 0.5271],
[ 0.6446],
[ 0.7222],
[ 0.5722],
[ 0.7676],
[ 0.6824],
[ 0.1981],
[ 0.8089],
[ 0.6296],
[ 0.6748],
[ 0.7515],
[ 0.5103],
[ 0.9052],
[ 0.8405],
[ 0.9092],
[ 0.6918],
[ 0.6477],
[ 0.5402],
[ 0.6477],
[ 0.4210],
[ 0.6973],
[ 0.6019],
[ 0.5364],
[ 0.8134],
[ 0.5607],
[ 0.7096],
[ 0.5894],
[ 0.3866],
[ 1.0600],
[ 0.7347],
[ 0.8129],
[ 1.2088],
[ 0.8825],
[ 0.7179],
[ 1.0115],
[ 0.7013],
[ 1.0128],
[ 0.9747],
[ 1.2759],
[ 0.7655],
[ 1.0094],
[ 0.7805],
[ 0.6091],
[ 1.2033],
[ 0.9678],
[ 0.8219],
[ 0.8157],
[ 0.9188],
[ 0.7436],
[ 0.8910],
[ 0.7291],
[ 0.9559],
[ 0.9389],
[ 1.2030],
[ 1.0495],
[ 1.1811],
[ 0.8884],
[ 0.8390],
[ 0.9894],
[ 0.9238],
[ 0.7628],
[ 0.5421],
[ 1.5147],
[ 0.6971],
[ 0.6740],
[ 0.8342],
[ 0.6554],
[ 0.7455],
[ 0.6916],
[ 1.2706],
[ 1.1277],
[ 0.9248],
[ 0.9976],
[ 1.2404],
[ 0.6919],
[ 1.3449],
[ 1.1243],
[ 1.0492],
[ 0.9266],
[ 1.1194],
[ 1.0304],
[ 1.1323],
[ 1.2372],
[ 0.8300],
[ 1.1916],
[ 1.0923],
[ 0.8313],
[ 0.8572],
[ 1.1128],
[ 1.0047],
[ 1.1544],
[ 0.9745],
[ 1.0503],
[ 0.9171],
[ 0.8073],
[ 1.2056],
[ 1.0976],
[ 0.9910],
[ 1.1834],
[ 1.1389],
[ 0.9142],
[ 0.9367],
[ 1.0121],
[ 0.7704],
[ 1.0558],
[ 0.7306],
[ 0.8117],
[ 0.7061],
[ 1.2315],
[ 0.9015],
[ 0.9339],
[ 0.5016],
[ 0.9227],
[ 1.2568],
[ 0.9444],
[ 1.1198],
[ 0.9431],
[ 1.0997],
[ 1.3078],
[ 0.8336],
[ 1.2692],
[ 0.8424],
[ 0.8702],
[ 1.4820],
[ 1.3248],
[ 0.9324],
[ 0.6538],
[ 1.2011],
[ 1.0170],
[ 0.7863],
[ 1.0178],
[ 0.6519],
[ 0.5970],
[ 0.9052],
[ 0.6846],
[ 0.7737],
[ 0.9104],
[ 0.8439],
[ 1.0066],
[ 1.0787],
[ 0.9661],
[ 0.9923],
[ 0.7922],
[ 0.8316],
[ 0.9553],
[ 0.9952],
[ 0.8680],
[ 1.1226],
[ 0.8213],
[ 0.9151],
[ 0.7748],
[ 0.9953],
[ 0.7773],
[ 0.7916],
[ 0.7321],
[ 0.9130],
[ 1.1433],
[ 0.7060],
[ 0.8066],
[ 0.8709],
[ 0.7426],
[ 0.8718],
[ 1.0973],
[ 0.7097],
[ 0.9438],
[ 0.8164],
[ 0.8013],
[ 0.6236],
[ 0.7180],
[ 0.9188],
[ 0.8016],
[ 0.9741],
[ 0.6271],
[ 0.5747],
[ 0.8007],
[ 0.7754],
[ 0.4877],
[ 0.4746],
[ 0.8654],
[ 0.4743],
[ 0.9015],
[ 0.8082],
[ 0.5449],
[ 0.9299],
[ 0.2003],
[ 0.5466],
[ 0.4355],
[ 0.7900],
[ 0.4343],
[ 0.7224],
[ 0.8585],
[ 0.5714],
[ 0.5306],
[ 0.6594],
[ 0.0640],
[ 0.3203],
[ 0.5463],
[ 0.5048],
[ 0.1935],
[ 0.2883],
[ 0.6778],
[ 0.5014],
[ 0.5235],
[ 0.5718],
[ 0.4587],
[ 0.2808],
[ 0.4073],
[ 0.8632],
[ 0.8862],
[ 0.5757],
[ 0.3372],
[ 0.2566],
[ 0.7858],
[ 0.3713],
[ 0.1589],
[ 0.3243],
[ 0.4270],
[ 0.0565],
[ 0.2885],
[ 0.3257],
[ 0.2196],
[ 0.3159],
[ 0.2361],
[ 0.1087],
[ 0.2224],
[ 0.2633],
[ 0.5037],
[ 0.1980],
[ 0.1530],
[ 0.2780],
[-0.1399],
[ 0.5331],
[ 0.3530],
[ 0.3342],
[ 0.2098],
[-0.0165],
[ 0.1318],
[ 0.4510],
[-0.1959],
[ 0.0966],
[ 0.0789],
[ 0.3381],
[-0.1917],
[ 0.1518],
[ 0.3640],
[ 0.0956],
[ 0.2535],
[-0.3988],
[-0.3479],
[ 0.3864],
[-0.2639],
[-0.2368],
[ 0.0258],
[ 0.2441],
[ 0.0687],
[ 0.0457],
[ 0.2286],
[-0.0947],
[-0.1189],
[ 0.1360],
[-0.0990],
[-0.2447],
[ 0.2135],
[-0.1830],
[-0.4583],
[-0.1795],
[-0.1361],
[-0.0553],
[-0.2864],
[-0.2307],
[-0.4651],
[-0.1889],
[-0.3185],
[-0.5318],
[-0.3012],
[ 0.0062],
[ 0.1046],
[-0.2321],
[-0.2945],
[-0.0242],
[-0.0586],
[-0.2307],
[-0.2479],
[-0.0382],
[-0.1509],
[-0.5055],
[-0.3759],
[ 0.2139],
[-0.2129],
[-0.3605],
[-0.5222],
[-0.6530],
[-0.6716],
[-0.4330],
[-0.2577],
[-0.2672],
[-0.1297],
[-0.9203],
[-0.5832],
[-0.2640],
[-0.4996],
[-0.2625],
[-0.4407],
[-0.8864],
[-0.2508],
[-0.4827],
[-0.3131],
[-0.2570],
[-0.7116],
[-0.5357],
[-0.7074]]))In [86]:
batch_size, n_train = 16, 600
# 只有前n_train个样本用于训练
train_iter = d2l.load_array((features[:n_train], labels[:n_train]),
batch_size, is_train=True)
In [87]:
def init_weights(m):
if type(m) == nn.Linear:
nn.init.xavier_uniform_(m.weight)In [88]:
def get_net():
net = nn.Sequential(nn.Linear(4, 10),
nn.ReLU(),
nn.Linear(10, 1))
net.apply(init_weights)
return net
loss = nn.MSELoss(reduction='none')In [89]:
def train(net, train_iter, loss, epochs, lr):
trainer = torch.optim.Adam(net.parameters(), lr)
for epoch in range(epochs):
for X, y in train_iter:
trainer.zero_grad()
l = loss(net(X), y)
l.sum().backward()
trainer.step()
print(f'epoch {epoch + 1}, '
f'loss: {d2l.evaluate_loss(net, train_iter, loss):f}')
net = get_net()
train(net, train_iter, loss, 5, 0.01)epoch 1, loss: 0.082748 epoch 2, loss: 0.066498 epoch 3, loss: 0.061828 epoch 4, loss: 0.059193 epoch 5, loss: 0.058498
/home/yukun/.conda/envs/nn/lib/python3.11/site-packages/d2l/torch.py:3179: UserWarning: Converting a tensor with requires_grad=True to a scalar may lead to unexpected behavior. Consider using tensor.detach() first. (Triggered internally at /pytorch/torch/csrc/autograd/generated/python_variable_methods.cpp:836.) self.data = [a + float(b) for a, b in zip(self.data, args)]
In [90]:
onestep_preds = net(features)
d2l.plot([time, time[tau:]],
[x.detach().numpy(), onestep_preds.detach().numpy()], 'time',
'x', legend=['data', '1-step preds'], xlim=[1, 1000],
figsize=(6, 3))
In [91]:
multistep_preds = torch.zeros(T)
multistep_preds[: n_train + tau] = x[: n_train + tau]
for i in range(n_train + tau, T):
multistep_preds[i] = net(
multistep_preds[i - tau:i].reshape((1, -1)))
d2l.plot([time, time[tau:], time[n_train + tau:]],
[x.detach().numpy(), onestep_preds.detach().numpy(),
multistep_preds[n_train + tau:].detach().numpy()], 'time',
'x', legend=['data', '1-step preds', 'multistep preds'],
xlim=[1, 1000], figsize=(6, 3))
In [92]:
import collections
import reIn [93]:
d2l.DATA_HUB['time_machine'] = (d2l.DATA_URL + 'timemachine.txt',
'090b5e7e70c295757f55df93cb0a180b9691891a')
def read_time_machine(): #@save
"""将时间机器数据集加载到文本行的列表中"""
with open(d2l.download('time_machine'), 'r') as f:
lines = f.readlines()
return [re.sub('[^A-Za-z]+', ' ', line).strip().lower() for line in lines]
lines = read_time_machine()
print(f'# 文本总行数: {len(lines)}')
print(lines[0])
print(lines[10])# 文本总行数: 3221 the time machine by h g wells twinkled and his usually pale face was flushed and animated the
In [94]:
def tokenize(lines, token='word'): #@save
"""将文本行拆分为单词或字符词元"""
if token == 'word':
return [line.split() for line in lines]
elif token == 'char':
return [list(line) for line in lines]
else:
print('错误：未知词元类型：' + token)
tokens = tokenize(lines)
for i in range(11):
print(tokens[i])['the', 'time', 'machine', 'by', 'h', 'g', 'wells'] [] [] [] [] ['i'] [] [] ['the', 'time', 'traveller', 'for', 'so', 'it', 'will', 'be', 'convenient', 'to', 'speak', 'of', 'him'] ['was', 'expounding', 'a', 'recondite', 'matter', 'to', 'us', 'his', 'grey', 'eyes', 'shone', 'and'] ['twinkled', 'and', 'his', 'usually', 'pale', 'face', 'was', 'flushed', 'and', 'animated', 'the']
In [95]:
def count_corpus(tokens): #@save
"""统计词元的频率"""
# 这里的tokens是1D列表或2D列表
if len(tokens) == 0 or isinstance(tokens[0], list):
# 将词元列表展平成一个列表
tokens = [token for line in tokens for token in line]
return collections.Counter(tokens)
class Vocab: #@save
"""文本词表"""
def __init__(self, tokens=None, min_freq=0, reserved_tokens=None):
if tokens is None:
tokens = []
if reserved_tokens is None:
reserved_tokens = []
# 按出现频率排序
counter = count_corpus(tokens)
self._token_freqs = sorted(counter.items(), key=lambda x: x[1],
reverse=True)
# 未知词元的索引为0
self.idx_to_token = ['<unk>'] + reserved_tokens
self.token_to_idx = {token: idx
for idx, token in enumerate(self.idx_to_token)}
for token, freq in self._token_freqs:
if freq < min_freq:
break
if token not in self.token_to_idx:
self.idx_to_token.append(token)
self.token_to_idx[token] = len(self.idx_to_token) - 1
def __len__(self):
return len(self.idx_to_token)
def __getitem__(self, tokens):
if not isinstance(tokens, (list, tuple)):
return self.token_to_idx.get(tokens, self.unk)
return [self.__getitem__(token) for token in tokens]
def to_tokens(self, indices):
if not isinstance(indices, (list, tuple)):
return self.idx_to_token[indices]
return [self.idx_to_token[index] for index in indices]
@property
def unk(self): # 未知词元的索引为0
return 0
@property
def token_freqs(self):
return self._token_freqsIn [96]:
vocab = Vocab(tokens)
print(list(vocab.token_to_idx.items())[:10])[('<unk>', 0), ('the', 1), ('i', 2), ('and', 3), ('of', 4), ('a', 5), ('to', 6), ('was', 7), ('in', 8), ('that', 9)]
In [97]:
for i in [0, 100]:
print('文本:', tokens[i])
print('索引:', vocab[tokens[i]])文本: ['the', 'time', 'machine', 'by', 'h', 'g', 'wells'] 索引: [1, 19, 50, 40, 2183, 2184, 400] 文本: ['were', 'three', 'dimensional', 'representations', 'of', 'his', 'four', 'dimensioned'] 索引: [20, 175, 1452, 2250, 4, 25, 262, 2251]
In [98]:
def load_corpus_time_machine(max_tokens=-1): #@save
"""返回时光机器数据集的词元索引列表和词表"""
lines = read_time_machine()
tokens = tokenize(lines, 'char')
vocab = Vocab(tokens)
# 因为时光机器数据集中的每个文本行不一定是一个句子或一个段落，
# 所以将所有文本行展平到一个列表中
corpus = [vocab[token] for line in tokens for token in line]
if max_tokens > 0:
corpus = corpus[:max_tokens]
return corpus, vocab
corpus, vocab = load_corpus_time_machine()
len(corpus), len(vocab)
Out [98]:
(170580, 28)
In [99]:
tokens = d2l.tokenize(read_time_machine())
# 因为每个文本行不一定是一个句子或一个段落，因此我们把所有文本行拼接到一起
corpus = [token for line in tokens for token in line]
vocab = d2l.Vocab(corpus)
vocab.token_freqs[:10]Out [99]:
[('the', 2261),
('i', 1267),
('and', 1245),
('of', 1155),
('a', 816),
('to', 695),
('was', 552),
('in', 541),
('that', 443),
('my', 440)]In [100]:
freqs = [freq for token, freq in vocab.token_freqs]
d2l.plot(freqs, xlabel='token: x', ylabel='frequency: n(x)',
xscale='log', yscale='log')
In [101]:
bigram_tokens = [pair for pair in zip(corpus[:-1], corpus[1:])]
bigram_vocab = Vocab(bigram_tokens)
bigram_vocab.token_freqs[:10]Out [101]:
[(('of', 'the'), 309),
(('in', 'the'), 169),
(('i', 'had'), 130),
(('i', 'was'), 112),
(('and', 'the'), 109),
(('the', 'time'), 102),
(('it', 'was'), 99),
(('to', 'the'), 85),
(('as', 'i'), 78),
(('of', 'a'), 73)]Warning:
Output truncated. This notebook contains too many cells to display efficiently.