我对PyTorch dataloader里的shuffle=True的理解

时间:2022-11-20 11:37:19

对shuffle=true的理解:

之前不了解shuffle的实际效果,假设有数据a,b,c,d,不知道batch_size=2后打乱,具体是如下哪一种情况:

1.先按顺序取batch,对batch内打乱,即先取a,b,a,b进行打乱;

2.先打乱,再取batch。

证明是第二种

?
1
2
3
4
shuffle (bool, optional): set to ``true`` to have the data reshuffled
at every epoch (default: ``false``).
if shuffle:
    sampler = randomsampler(dataset) #此时得到的是索引

补充:简单测试一下pytorch dataloader里的shuffle=true是如何工作的

看代码吧~

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
import sys
import torch
import random
import argparse
import numpy as np
import pandas as pd
import torch.nn as nn
from torch.nn import functional as f
from torch.optim import lr_scheduler
from torchvision import datasets, transforms
from torch.utils.data import tensordataset, dataloader, dataset
 
class dealdataset(dataset):
    def __init__(self):
        xy = np.loadtxt(open('./iris.csv','rb'), delimiter=',', dtype=np.float32)
        #data = pd.read_csv("iris.csv",header=none)
        #xy = data.values
        self.x_data = torch.from_numpy(xy[:, 0:-1])
        self.y_data = torch.from_numpy(xy[:, [-1]])
        self.len = xy.shape[0]
    
    def __getitem__(self, index):
        return self.x_data[index], self.y_data[index]
 
    def __len__(self):
        return self.len
   
dealdataset = dealdataset()
train_loader2 = dataloader(dataset=dealdataset,
                          batch_size=2,
                          shuffle=true)
#print(dealdataset.x_data)
for i, data in enumerate(train_loader2):
    inputs, labels = data
 
    #inputs, labels = variable(inputs), variable(labels)
    print(inputs)
    #print("epoch:", epoch, "的第" , i, "个inputs", inputs.data.size(), "labels", labels.data.size())

简易数据集

我对PyTorch dataloader里的shuffle=True的理解我对PyTorch dataloader里的shuffle=True的理解

shuffle之后的结果,每次都是随机打乱,然后分成大小为n的若干个mini-batch.

我对PyTorch dataloader里的shuffle=True的理解

以上为个人经验,希望能给大家一个参考,也希望大家多多支持服务器之家。

原文链接:https://blog.csdn.net/qq_35248792/article/details/109510917