程序小记

记录一下，如何在torch中进行单个batch计算

batch_size = x.shape[0]
seq_len = x.shape[1] 
emo_induce = emo_induce.cpu().numpy() #aspect开始和结束的位置
text_len = x_len.cpu().numpy() #传入子句的真实长度

emo_weight = [[] for i in range(batch_size)]
context_weight = [[] for i in range(batch_size)]

for i in range(batch_size):
    # print('context_len = ', context_len)
    for j in range(emo_induce[i,0]):
        context_weight[i].append(1)
        emo_weight[i].append(0)
    for j in range(emo_induce[i,0], emo_induce[i,1]):
        context_weight[i].append(0)
        emo_weight[i].append(1)
    for j in range(emo_induce[i,1], text_len[i]):
        context_weight[i].append(1)
        emo_weight[i].append(0)
    for j in range(text_len[i], seq_len):
        context_weight[i].append(0)
        emo_weight[i].append(0)
context_weight = torch.tensor(context_weight).cuda()

** 判断符合列表中某些数字是否符合条件**

import torch
a =torch.Tensor([1,2,4,4,5])
print(torch.masked_select(a, a<4))

在linux终端下载github代码

git clone 网址

*.flat的用法

import numpy as np
a = np.arange(4).reshape(2,2)
print(a)
b = a.flat
print('b = ', b)
for i in b:
    print(i)
#迭代器可以用list进行输出
print(list(a.flat))
print(type(a.flat))#返回类型为 numpy.flatiter
#可以用索引对迭代器进行引号
a.flat[3]