PyTorch Load data set

1、Dataset： Data sets （ Support data indexing ）;

2、DataLoader： Load data set ;

3、Batch： Load all the data ;（ advantage ： Can maximize the use of vector computing optimization to improve the speed of Computing ; shortcoming ）

4、 Stochastic gradient descent ： A sample data ;（ advantage ： Get a good randomness , Overcome saddle point problem ; shortcoming ： The trained model is better than others , But because there is only one sample at a time , Can't use CPU/GPU Parallel ability , Optimization time is too long ）

5、 Code ：

import torch
import numpy as np
from torch.utils.data import Dataset, DataLoader    # Dataset： abstract class , Can't instantiate 
from torch.utils.data import DataLoader

#####################1  Get ready 、 Load data set #################################
# DiabetesDataset  Inherit  Dataset  The basic functions of the secondary class 
class DiabetesDataset(Dataset):
    def __init__(self, filepath):         #  initialization , According to the path （filepath） Or load simple data and tags 
        xy = np.loadtxt(filepath, delimiter=',', dtype=np.float32)
        self.len = xy.shape[0]
        self.x_data = torch.from_numpy(xy[ : , : -1])    #  All rows in the column after rent 
        self.y_data = torch.from_numpy(xy[ : , [-1]])    #  All rows in the last column 

#  In order to be in DiabetesDataset  After instantiation   Indexes dataset[index]  Extract the data 
def __getitem__(self, index):
    return self.x_data[index], self.y_data[index]

#  Quantity of data 
def __len__(self):
    return self.len

dataset = DiabetesDataset(“diabetes.csv.gz”)    #  Here  diabetes.csv.gz  No, 
train_loader = DataLoader(dataset = dataset,
                          batch_size= = 32,
                          shuffle = True,
                          num_workers = 2)

#########################2  Use class design model ###############################
class Model(torch.nn.Module):
    def __init__(self):
        super(Model, self).__init__()
        self.linear1 = torch.nn.Linear(8,6)
        self.linear1 = torch.nn.Linear(6,4)
        self.linear1 = torch.nn.Linear(4,1)
        self.sigmoid = torch.nn.Sigmoid()

    def forward(self, x):
        x = self.sigmoid(self.linear1(x))
        x = self.sigmoid(self.linear2(x))
        x = self.sigmoid(self.linear3(x))
        return x

model = Model()
###################3  Build loss function 、 Optimizer ###############################
criterion = torch.nn.BCELoss(size_average=False)          # BCE Loss 
optimizer = torch.optim.SGD(model.parameters(), lr=0.1)   #  Parameter optimization 

#####################4  Cycle training  #########################
for epoch in range(100):
    for i, data in enumerate(train_loader, 0):
        #  Prepare the data 
        inputs, labels = data
        #  Forward propagation 
        y_pred = model(inputs)
        loss = criterion(y_pred, labels)
        print(epoch, i, loss.item())
        #  Back propagation 
        optimizer.zero_grad()
        loss.backward()
        #  Update weights 
        optimizer.step()