当前位置：网站首页>"Experimental Details" to realize the function interpretation in nucleus sampling

"Experimental Details" to realize the function interpretation in nucleus sampling

2022-08-07 22:13:00 【365年jhwzgo】

对之前的torch.tensor进行累加

import torch
x = torch.linspace(0, 5, 6)
print(x)
y = torch.cumsum(x, dim=0)  
print(y)

在这里插入图片描述

import torch
x = torch.linspace(0, 5, 6).view(2, 3)
print(x)
y = torch.cumsum(x, dim=0)  
print(y)
z = torch.cumsum(x, dim=1)
print(z)

在这里插入图片描述
当p=3时

import torch
x = torch.linspace(0, 5, 6)
print(x)
sorted, sorted_indices = torch.sort(x, descending=True)
print(sorted, sorted_indices)

在这里插入图片描述

x = torch.tensor([2, 2, 5, 3, 1, 4],dtype=torch.float64)
x[..., 1:] = x[..., :-1].clone()
x

在这里插入图片描述
当对一个tensorwhen operating at the same time,A new memory needs to be cloned

版权声明
本文为[365年jhwzgo]所创，转载请带上原文链接，感谢
https://yzsam.com/2022/219/202208072206249416.html