- 博客(48)
- 收藏
- 关注
原创 qa问答机器人pysparnn问题的召回
"""构造召回的模型"""from sklearn.feature_extraction.text import TfidfVectorizerimport pysparnn.cluster_index as cifrom cut_sentence import cutimport jsondef prepar_recall_datas(): qa_dict =...
2020-02-29 21:38:00 339
原创 pysparnn 模块使用,相似句子召回
import pysparnn.cluster_index as cifrom sklearn.feature_extraction.text import TfidfVectorizerdata = [ "hello world", "oh hello there", "Play it", "Play it again Sam",]tv ...
2020-02-29 21:16:00 335
原创 pytorch seq2seq闲聊机器人beam search返回结果
decoder.py"""实现解码器"""import heapqimport torch.nn as nnimport configimport torchimport torch.nn.functional as Fimport numpy as npimport randomfrom chatbot.attention import Attention...
2020-02-29 21:15:00 679
原创 pytorch seq2seq闲聊机器人加入attention机制
attention.py"""实现attention"""import torchimport torch.nn as nnimport torch.nn.functional as Fimport configclass Attention(nn.Module): def __init__(self,method="general"): s...
2020-02-28 21:36:00 226
原创 python 中自带的堆模块heapq
import heapqmy_heap = [] #使用列表保存数据#网列表中插入数据,优先级使用插入的内容来表示,就是一个比较大小的操作,越大优先级越高heapq.heappush(my_heap,[29,True,"xiaohong","asdfa"])heapq.heappush(my_heap,[28,False,"xiaowang"])heapq.heappu...
2020-02-28 20:44:00 186
原创 pytorch seq2seq闲聊机器人
cut_sentence.py"""实现句子的分词注意点:1. 实现单个字分词2. 实现按照词语分词2.1 加载词典3. 使用停用词"""import stringimport jiebaimport jieba.posseg as psgimport loggingstopwords_path = "../corpus/stopwords.txt"...
2020-02-28 16:17:00 442
原创 pytorch 矩阵数据增加维度unsqueeze和降低维度squeeze
增加一个维度 out.unsqueeze(-1)降低一个维度 out.squeeze(dim=1)
2020-02-23 11:32:00 4122
原创 pytorch seq2seq模型中加入teacher_forcing机制
在循环内加的teacher forcing机制,这种为目标确定的时候,可以这样加。目标不确定,需要在循环外加。decoder.py 中的修改"""实现解码器"""import torch.nn as nnimport configimport torchimport torch.nn.functional as Fimport numpy as npimport ...
2020-02-22 00:11:00 2288 1
原创 pytorch seq2seq模型训练测试
num_sequence.py"""数字序列化方法"""class NumSequence: """ input : intintint output :[int,int,int] """ PAD_TAG = "<PAD>" UNK_TAG = "<UNK>" SOS_TAG = "<SOS...
2020-02-20 21:07:00 888
原创 python os模块判断文件是否存在,file_path获取当前文件路径
import osfile_path = os.path.dirname(__file__)os.path.exists(test_file.txt)
2020-02-20 20:53:00 189
原创 fasttext的使用,预料格式,调用方法
数据格式:分词后的句子+\t__label__+标签fasttext_model.pyfrom fasttext import FastTextimport numpy as npdef get_data_path(by_word=True,train=True): if by_word: return "./classify/data_by_wor...
2020-02-20 00:05:00 345
原创 python中文语料分词处理,按字或者词cut_sentence
cut_sentence.pyimport stringimport jiebaimport jieba.posseg as psgimport logging#关闭jieba日制jieba.setLogLevel(logging.INFO)jieba.load_userdict("./corpus/keywords.txt")stopwords_path = "....
2020-02-19 22:43:00 16338 2
原创 pytorch LSTM情感分类全部代码
先运行main.py进行文本序列化,再train.py模型训练dataset.pyfrom torch.utils.data import DataLoader,Datasetimport torchimport osfrom utils import tokenlizeimport configclass ImdbDataset(Dataset): d...
2020-02-18 10:15:00 2340 2
原创 pytorch 文本情感分类和命名实体识别NER中LSTM输出的区别
文本情感分类:文本情感分类采用LSTM的最后一层输出比如双层的LSTM,使用正向的最后一层和反向的最后一层进行拼接 def forward(self,input): ''' :param input: :return: ''' input_embeded = self.embedding...
2020-02-18 10:09:00 574
原创 windows 系统查看NVIDIA显卡GPU情况,nvidia-smi在windows上使用
cd C:\Program Files\NVIDIA Corporation\NVSMInvidia-smi
2020-02-17 23:12:00 285
原创 pytorch 中LSTM模型获取最后一层的输出结果,单向或双向
单向LSTMimport torch.nn as nnimport torchseq_len = 20batch_size = 64embedding_dim = 100num_embeddings = 300hidden_size = 128number_layer = 3input = torch.randint(low=0,high=256,size=[...
2020-02-17 18:31:00 5513 3
原创 文本序列化【通用】word2sequence,文本序列字典保存
'''文本序列化'''class WordSequence(): UNK_TAG = "<UNK>" PAD_TAG = "<PAD>" UNK = 1 PAD = 0 def __init__(self): self.dict = { self.UNK_TAG:self....
2020-02-16 22:56:00 1150
原创 英文语料的分词处理
"""实现额外的方法"""import redef tokenlize(sentence): """ 进行文本分词 :param sentence: str :return: [str,str,str] """ fileters = ['!', '"', '#', '$', '%', '&', '\(', '\)', ...
2020-02-16 21:15:00 446
原创 pytorch GPU训练好的模型使用CPU加载
torch.load('tensors.pt')# 把所有的张量加载到CPU中torch.load('tensors.pt', map_location=lambda storage, loc: storage)# 把所有的张量加载到GPU 1中torch.load('tensors.pt', map_location=lambda storage, loc: storage....
2020-02-16 15:04:00 937
原创 pytorch 手写数字识别项目 增量式训练
dataset.py'''准备数据集'''import torchfrom torch.utils.data import DataLoaderfrom torchvision.datasets import MNISTfrom torchvision.transforms import ToTensor,Compose,Normalizeimport torchv...
2020-02-15 22:38:00 942 2
原创 pytorch 中模型的保存与加载,增量训练
让模型接着上次保存好的模型训练,模型加载#实例化模型、优化器、损失函数model = MnistModel().to(config.device)optimizer = optim.Adam(model.parameters(),lr=0.01)if os.path.exists("./model/mnist_net.pt"): model.load_state_...
2020-02-15 21:42:00 1714
原创 pytorch 中HWC转CHW
import torchimport numpy as npfrom torchvision.transforms import ToTensort = torch.tensor(np.arange(24).reshape(2,4,3))print(t)#HWC 转CHWprint(t.transpose(0,2).transpose(1,2))print(t.pe...
2020-02-15 18:07:00 5929 2
原创 pytorch torchversion自带的数据集
from torchvision.datasets import MNIST# import torchvision# torchvision.datasets.#准备数据集mnist = MNIST(root="./mnist",train=True,download=True)print(mnist)mnist[0][0].show()print(len...
2020-02-15 17:32:00 1132
原创 pytorch Dataset数据集和Dataloader迭代数据集
import torchfrom torch.utils.data import Dataset,DataLoaderclass SmsDataset(Dataset): def __init__(self): self.file_path = "./SMSSpamCollection" self.lines = open(self.fi...
2020-02-14 16:42:00 560
原创 pytorch Model Linear实现线性回归CUDA版本
实验代码import torchimport torch.nn as nn#y = wx + bclass MyModel(nn.Module): def __init__(self): super(MyModel,self).__init__() #自定义代码 # self.w = torch.rand([500,...
2020-02-12 19:34:00 699
原创 pytorch实现手动线性回归
import torchimport matplotlib.pyplot as pltlearning_rate = 0.1#准备数据 #y = 3x +0.8x = torch.randn([500,1])y_true = 3*x + 0.8#计算预测值w = torch.rand([],requires_grad=True)b = torch.tensor(0,...
2020-02-12 17:37:00 229
原创 pytorch中的前项计算和反向传播
前项计算1import torch# (3*(x+2)^2)/4#grad_fn 保留计算的过程x = torch.ones([2,2],requires_grad=True)print(x)y = x+2print(y)z = 3*y.pow(2)print(z)out = z.mean()print(out)#带有反向传播属性的tensor不能直接...
2020-02-12 16:56:00 396
原创 pytorch中CUDA类型的转换
import torchimport numpy as npdevice = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")x = torch.tensor(np.arange(15).reshape(3,5))if torch.cuda.is_available(): device =...
2020-02-12 16:11:00 877
原创 pytorch中tensor的属性 类型转换 形状变换 转置 最大值
import torchimport numpy as npa = torch.tensor([[[1]]])#只有一个数据的时候,获取其数值print(a.item())#tensor转化为nparrayb = a.numpy()print(b,type(b),type(a))#获取张量的形状a = torch.tensor(np.arange(30).resh...
2020-02-10 22:22:00 968
原创 pytorch中tensor张量的创建
import torchimport numpy as npprint(torch.tensor([1,2,3]))print(torch.tensor(np.arange(15).reshape(3,5)))print(torch.empty([3,4]))print(torch.ones([3,4]))print(torch.zeros([3,4]))#0-1之间的...
2020-02-10 21:51:00 99
原创 TensorFlow keras 迁移学习
数据的读取import tensorflow as tffrom tensorflow.python import kerasfrom tensorflow.python.keras.preprocessing.image import ImageDataGeneratorclass TransferModel(object)...
2020-02-06 00:06:00 635 1
原创 TensorFlow keras vgg16net的使用
from tensorflow.python.keras.applications.vgg16 import VGG16,preprocess_input,decode_predictionsfrom tensorflow.python.keras.preprocessing.image import load_img,img_to_arraydef predict(): ...
2020-02-03 20:57:00 678
空空如也
空空如也
TA创建的收藏夹 TA关注的收藏夹
TA关注的人