pytorch修改tensor数据类型

最新推荐文章于 2024-08-20 12:41:32 发布

原创最新推荐文章于 2024-08-20 12:41:32 发布 · 2.2w 阅读

30 ·

CC 4.0 BY-SA版权

本文为CSDN博主"呆萌的代Ma"原创文章，转载请注明博客链接：https://blog.csdn.net/weixin_35757704/

文章标签：

#pytorch

人工智能专栏收录该内容

28 篇文章

订阅专栏

本文介绍在PyTorch中如何对Tensor进行多种类型的转换，包括但不限于float32到long、half、int、double、char、byte及short类型的转换方法。通过两种不同方式展示这些操作的具体实现。

部署运行你感兴趣的模型镜像

官方类型如下：
在这里插入图片描述

方法一

import torch

my_tensor = torch.randn(2, 4)  # 默认为float32类型
print("原始：", my_tensor)
print('____________________________________________________________')

# torch.long() 将tensor投射为long类型
long_tensor = my_tensor.long()
print(long_tensor)
print('____________________________________________________________')

# torch.half()将tensor投射为半精度浮点类型
half_tensor = my_tensor.half()
print(half_tensor)
print('____________________________________________________________')

# torch.int()将该tensor投射为int类型
int_tensor = my_tensor.int()
print(int_tensor)
print('____________________________________________________________')

# torch.double()将该tensor投射为double类型
double_tensor = my_tensor.double()
print(double_tensor)
print('____________________________________________________________')

# torch.float()将该tensor投射为float类型
float_tensor = my_tensor.float()
print(float_tensor)
print('____________________________________________________________')

# torch.char()将该tensor投射为char类型
char_tensor = my_tensor.char()
print(char_tensor)
print('____________________________________________________________')

# torch.byte()将该tensor投射为byte类型
byte_tensor = my_tensor.byte()
print(byte_tensor)
print('____________________________________________________________')

# torch.short()将该tensor投射为short类型
short_tensor = my_tensor.short()
print(short_tensor)
print('____________________________________________________________')

方法二

my_tensor = torch.randn(2, 4)  # 默认为float32类型

my_tensor.type(torch.float16)
print('____________________________________________________________')

print(my_tensor.type(torch.float16))
print('____________________________________________________________')

print(my_tensor.type(torch.float32))
print('____________________________________________________________')

print(my_tensor.type(torch.int32))
print('____________________________________________________________')

print(my_tensor.type(torch.long))
print('____________________________________________________________')