10.18Pytorch学习日志

torch.max()

torch.matmul()：
tensor的乘法（2维就普通矩阵乘法）

1. 均匀分布
   torch.nn.init.uniform_(tensor, a=0, b=1)
   服从~U ( a , b ) U(a, b)U(a,b)

2. 正太分布
   torch.nn.init.normal_(tensor, mean=0, std=1)
   服从~N ( m e a n , s t d ) N(mean, std)N(mean,std)

3. 初始化为常数
   torch.nn.init.constant_(tensor, val)

---

把cpu转到gpu跑：
for X, y in train_iter:
x_train = torch.autograd.Variable(X.cuda())
y_train = torch.autograd.Variable(y.cuda())
net = LinearNet(num_inputs, num_outputs)
net = net.cuda()

import torch
from torch import nn
from torch.nn import init
import numpy as np
import sys

sys.path.append('..')
import d2lzh_pytorch as d2l

# 定义模型
num_inputs, num_outputs, num_hiddens = 784, 10, 256

net = nn.Sequential(
    d2l.FlattenLayer(),
    nn.Linear(num_inputs, num_hiddens),
    nn.ReLU(),
    nn.Linear(num_hiddens, num_outputs),
)

for params in net.parameters():
    init.normal_(params, mean=0, std=0.01)

# 读取数据， 训练模型
batch_size = 256
train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size)
loss = torch.nn.CrossEntropyLoss()
optimizer = torch.optim.SGD(net.parameters(),lr=0.5)

num_epochs = 5
d2l.train_ch3(net, train_iter, test_iter, loss, num_epochs,
              batch_size, None, None, optimizer)