jianshanzhange的博客

原创 django初识加扫清障碍(已放弃)

1.anaconda安装django和pymysql框架。conda install djangopip install pymysql成功情况下图所示：2.pycharm新建django文件。使用已有的anaconda环境，这里是主环境，可以随时切换。成功情况下图所示：3.运行django文件。在pycharm中的Terminal窗口输入运行指令。python manage.py runserver如果遇到以下报错，则将settings.py中的’DIRS’: [BASE

原创 文本和损失

# coding: UTF-8import torchimport torch.nn as nnfunc=nn.CrossEntropyLoss()a=torch.Tensor([[ 0.0606,0.1610,0.2990,0.2101, 0.5104], [0.6388,0.4053, 0.4196, 0.7060, 0.2793], [ 0.3973,0.6114, 0.1127, 0.7732, 0.0592]])b=[0.4

原创 更新加入LSTM的推特文本情感分类

import torchimport torchtextimport torch.nn as nnimport torch.nn.functional as Ffrom torchtext.vocab import GloVeimport numpy as npimport pandas as pdimport matplotlib.pyplot as pltimport refrom sklearn.model_selection import train_test_splitimpo

原创 基于pytorch的推特评论情感分类

import torchimport torchtextimport torch.nn as nnimport torch.nn.functional as Ffrom torchtext.vocab import GloVeimport numpy as npimport pandas as pdimport matplotlib.pyplot as pltimport refrom sklearn.model_selection import train_test_splitimpo

原创 基于pytorch的LSTM的简单文本分类

import torchimport torchtextimport numpy as npimport torch.nn as nnimport torch.nn.functional as Ffrom torchtext.vocab import GloVeimport timestart=time.time()#每篇提取200个单词TEXT = torchtext.data.Field(lower=True, fix_length=200, batch_first=False)L

原创 基于pytorch的GRUCell的简单文本分类

import torchimport torchtextimport numpy as npimport torch.nn as nnimport torch.nn.functional as Ffrom torchtext.vocab import GloVeimport timestart=time.time()#每篇提取200个单词TEXT = torchtext.data.Field(lower=True, fix_length=200, batch_first=False)L

原创 基于pytorch的LSTNCell的简单文本分类

import torchimport torchtextimport numpy as npimport torch.nn as nnimport torch.nn.functional as Ffrom torchtext.vocab import GloVeimport timestart=time.time()#每篇提取200个单词TEXT = torchtext.data.Field(lower=True, fix_length=200, batch_first=False)L

原创 LSTM、GRU相关基础知识

Xt-1可以理解为单词一：IXt-2可以理解为单词二：loveXt-3可以理解为单词三：you有两个状态传递给了后面的单元，上面的叫单元状态输出，代表了每一个单元的状态。单元里的ⓧ是对前面单元状态输出的选择，可以换让信息不通过、完全通过、通过一部分。单元里的⊕是将上一个单元传输过来的单元状态，添加进新的状态。下面的是代表是当前的隐藏单元的输出，输入Xt-2与前面单元的隐藏单元输出相结合，经过一系列运算得到新的隐藏单元输出。得到的新输出作为本单元的输出或者交给下一个单元格。...

原创 基于pytorch的RNNCell的简单文本分类(更新GPU使用方法)

import torchimport torchtextimport numpy as npimport torch.nn as nnimport torch.nn.functional as Ffrom torchtext.vocab import GloVeimport timestart=time.time()#每篇提取200个单词TEXT = torchtext.data.Field(lower=True, fix_length=200, batch_first=False)L

原创 RNN相关基础知识

原创 基于pytorch的预训练词向量的简单文本分类

import torchimport torchtextimport numpy as npimport torch.nn.functional as Ffrom torchtext.vocab import GloVeimport torch.nn as nnimport timefrom torchtext.vocab import GloVe#不推荐使用文本预训练词向量 这里仅写出使用方法start=time.time()TEXT = torchtext.data.Field

原创 基于pytorch的自建模型简单文本分类(fit函数有变化)

import torchimport torchtextimport numpy as npimport torch.nn.functional as Ffrom torchtext.vocab import GloVeimport torch.nn as nnimport timestart=time.time()#创建字段#小写 截取评论长度为200(少了填充、多了截取) 第一维度是batchTEXT = torchtext.data.Field(lower=True,fix_l

原创 自然语言处理相关基础知识

import torchimport numpy as npimport string#步骤:1、分词 2、创建词表 3、词嵌入表示s = 'Life is not easy for any of us. We must work, and above all we must believe in ourselves. We must believe that each one of us is able to do some thing well. And ...

原创 Inception相关基础知识

原创 基于pytorch的自己搭建残差神经网络

import torchfrom torch import nnimport torch.nn.functional as Fimport torchvisionclass ResnetbasicBlock(nn.Module): def __int__(self,in_channels,out_channels): super().__init__() self.conv1 == nn.Conv2d(in_channels,out_c...

原创 基于pytorch的训练、保存模型、调用模型的四种天气分类

import torchimport torch.nn as nnimport pandas as pdimport numpy as npimport matplotlib.pyplot as pltimport torch.nn.functional as Fimport torchvisionimport timeimport osimport shutilimport copyfrom torchvision import datasets,transformsfrom to

原创 微调与迁移学习相关基础知识

原创 基于pytorch的自动下载RESNET预训练模型的四种天气分类

import torchimport torch.nn as nnimport pandas as pdimport numpy as npimport matplotlib.pyplot as pltimport torch.nn.functional as Fimport torchvisionimport timeimport osimport shutilfrom torchvision import datasets,transformsfrom torch.utils.da

原创 基于pytorch的自动下载VGG16预训练模型含图像数据增强和学习速率衰减的四种天气分类

import torchimport torch.nn as nnimport pandas as pdimport numpy as npimport matplotlib.pyplot as pltimport torch.nn.functional as Fimport torchvisionimport timeimport osimport shutilfrom torchvision import datasets,transformsfrom torch.utils.da

原创 基于pytorch的自动下载VGG16预训练模型的四种天气分类

使用情况：model = torchvision.models.vgg16(pretrained=True)将训练好的卷积层参数冻结，抛弃训练好的线性层，然后使用自己搭建的好线性层。import torchimport torch.nn as nnimport pandas as pdimport numpy as npimport matplotlib.pyplot as pltimport torch.nn.functional as Fimport torchvisionimpo.

原创 预训练模型和VGG架构相关基础知识

原创 标准化和超参数相关基础知识

---------------------------------分割线-------------------------------------

原创 基于pytorch的四种天气分类含Dropout层和BN层

import torchimport torch.nn as nnimport pandas as pdimport numpy as npimport matplotlib.pyplot as pltimport torch.nn.functional as Fimport torchvisionimport timeimport osimport shutilfrom torchvision import datasets,transformsfrom torch.utils.da

原创 基于pytorch的自建卷积神经网络GPU的MNIST分类

import torchimport torch.nn as nnimport pandas as pdimport numpy as npimport matplotlib.pyplot as pltimport torch.nn.functional as Fimport torchvisionfrom torchvision import datasets,transformsfrom torch.utils.data import TensorDatasetfrom torch.u

原创 卷积神经网络相关基础知识

原创 基于pytorch的梯度下降、学习速率、反向传播算法和优化器

梯度下降算法用于寻找损失函数的极值点。 梯度的输出是一个由若干偏导数构成的向量，它的每个分量对应于函数对输入向量的相应分量的偏导。比如说有多个权重，一个输出。 梯度的输出向量表明了在每个位置损失函数增长最快的方向，可将它视为表示了在函数的每个位置向哪个方向移动函数值可以增长。 我们想让损失下降的最快，所以沿着反方向去移动。图解...

原创 基于pytorch的MNIST数据集分类

#利用fit()函数#首先,创建输入(dataloader)#然后,创建模型(model)#最后,创建损失函数import torchimport torch.nn as nnimport pandas as pdimport numpy as npimport matplotlib.pyplot as pltimport torch.nn.functional as Fimport torchvisionfrom torchvision import datasets,transfo

原创 基于pytorch的通用训练函数(待校验)

#编写fit()函数，参数为训练次数(epoch)、输入模型(model)、输入数据(train_dl,test_dl)#功能是对数据输入在模型上训练，并且返回loss和acc变化def fit(epoch,model,trainloader,testloder): correct = 0 total = 0 running_loss = 0 for x,y in trainloader: y_pred = model(x) loss =

原创 基于pytorch的部分鸢尾花数据集多项分类

import torchimport torch.nn as nnimport pandas as pdimport numpy as npimport matplotlib.pyplot as pltimport torch.nn.functional as Ffrom torch.utils.data import TensorDatasetfrom torch.utils.data import DataLoaderfrom sklearn.model_selection import

原创 基于pytorch多层感知器预测是否离职

import torchimport torch.nn as nnimport pandas as pdimport numpy as npimport matplotlib.pyplot as pltimport torch.nn.functional as Fdata=pd.read_csv(r'D:/pycharmworkspace/ISLR-master/HR.csv')#get_dummies将data里的salary数据置换成独热编码#data.join将转换成独热编码的数据

原创 基于pytorch的逻辑回归实例

import torchimport pandas as pdimport torch.nn as nnimport numpy as npimport matplotlib.pyplot as pltdata=pd.read_csv(r'D:/pycharmworkspace/ISLR-master/credit-a.csv',header=None)X = data.iloc[:,:-1]Y = data.iloc[:,-1].replace(-1,0)#数据预处理X = torc

原创 基于pytorch的线性回归预测

#f(x)=w*x+b x代表学历 f(x)代表收入#找到合适的w和b，使得(f(x)-y)^2越小越好 即求解参数w和bimport torchimport torch.nn as nnimport pandas as pdimport numpy as npimport matplotlib.pyplot as pltdata=pd.read_csv(r'D:/pycharmworkspace/ISLR-master/Income1.csv')plt.scatter(data.Edu

原创 基于pytorch详解VGG网络

如题所言 如下所示import torch.nn as nnimport torchclass VGG(nn.Module): def __init__(self, features, num_classes=1000, init_weights=False): #self 由make_features生成的特征 分类类别个数 是否对网络权重初始化 super(VGG, self).__init__() self.features = features #传