1- 张量

2024-09-01

1- 张量

1.1 创建张量

1）根据列表

import torch

data = [[1,2],[3,4]]
x_data = torch.tensor(data)

print(type(x_data))

2）根据numpy数组

import torch
import numpy as np

data = np.array([[1,2],[3,4]])
x_data = torch.from_numpy(data)

print(type(x_data))

3）根据Tensor

import torch
data = [[1,2],[3,4]]
x_data = torch.tensor(data)

x_ones = torch.ones_like(x_data)
x_zeros = torch.zeros_like(x_data)
x_rand = torch.rand_like(x_data, dtype=torch.float) #rand默认生成浮点数

print(f"张量1： {x_ones}")
print(f"张量0： {x_zeros}")
print(f"张量随机： {x_rand}")

4）根据形状

shape=(2,3)

x_ones = torch.ones(shape)
x_zeros = torch.zeros(shape)
x_rand = torch.rand(shape) #rand默认生成浮点数

print(f"张量1： {x_ones}")
print(f"张量0： {x_zeros}")
print(f"张量随机： {x_rand}")

1.2 张量的属性

属性	解释
shape	形状
dtype	数据类型
device	在哪个设备上

1.3 张量操作

1.3.1 张量判断

操作	解释
is_tensor()	是否为tensor
is_complex()	元素为复数
is_floating_point()	元素为浮点数
numel()	元素个数

1.3.2 创建操作

操作	解释
tensor()	根据内容创建张量
zeros() ones() eye() empty() full(形状, 数值)	全0矩阵全1矩阵单位矩阵未初始化的张量单元素填充
zeros_like() ones_like()	全0矩阵全1矩阵
arange()	范围：[ )
range()	范围：[ ]，steps指定步长
linspace() logspace	范围：[ ]，steps指定个数确定底数，生成一系列指数

1.3.3 索引，切片，拼接，变异

操作	解释
cat(tensors , dim=0)	按维度拼接（拼谁谁变长，其他维度相等）
stack(tensors , dim=0)	按维度堆叠（会新增维度）
chunk(input, chunks, dim=0)	平均切成chunks段
dsplit(input, indices_or_sections) hsplit() split()	按维度2（深度）切割按维度1（高度）切割切割
dstack(tensors, , out=None*) hstack()	按维度3叠加（新增维度）按维度2拼接
reshape(input, shape)	改变形状
squeeze(input, dim=None)	压缩值为1的维度

torch.tensor(data, ***, dtype=None, device=None, requires_grad=False, pin_memory=False) → Tensor

torch.zeros(*size, ***, out=None, dtype=None, layout=torch.strided, device=None, requires_grad=False) → Tensor

import torch

x = torch.zeros(2,3)

print(x)

torch.zeros_like(input, ***, dtype=None, layout=None, device=None, requires_grad=False, memory_format=torch.preserve_format) → Tensor

import torch

x = torch.ones(2,3)
y = torch.zeros_like(x)

print(y)

torch.arange(start=0, end, step=1, ***, out=None, dtype=None, layout=torch.strided, device=None, requires_grad=False) → Tensor

import torch

x = torch.arange(5)

print(x)

torch.range(start=0, end, step=1, ***, out=None, dtype=None, layout=torch.strided, device=None, requires_grad=False) → Tensor

import torch

x = torch.range(start=0, end=5)

print(x)

torch.linspace(start, end, steps, ***, out=None, dtype=None, layout=torch.strided, device=None, requires_grad=False) → Tensor

import torch

x = torch.linspace(start=0, end=5,steps=3)

print(x)

torch.logspace(start, end, steps, base=10.0, ***, out=None, dtype=None, layout=torch.strided, device=None, requires_grad=False) → Tensor

import torch

x = torch.logspace(start=-10, end=10, steps=5)# 底数默认10

print(x)

torch.cat(tensors, dim=0, ***, out=None) → Tensor

import torch

x = torch.randn(2, 3)
x = torch.cat((x, x, x), 0)

print(x)

torch.dsplit(input, indices_or_sections) → List of Tensors

import torch
x = torch.arange(16.0).reshape(2, 2, 4)

# 按维度2均分成两个：[2,2,4] -> [2,2,2] * 2
x1 = torch.dsplit(x, 2)
# 按维度2分别占到[0,3），[3，6）[6,...)
# [2,2,4] -> [2,2,3] [2,2,1] [2,2,0]
x2 = torch.dsplit(x, [3, 6])

print(x1)
print(x2)

torch.stack(tensors, dim=0, ***, out=None) → Tensor

import torch
x1 = torch.randn(2, 3)
x2 = torch.ones(2,3)

y1 = torch.stack((x1, x2))
y2 = torch.stack((x1, x2), dim=1)
y3 = torch.stack((x1, x2), dim=2)

print(y1)
print(y2)
print(y3)

torch.dstack(tensors, ***, out=None) → Tensor

import torch
x1 = torch.tensor([1, 2, 3])
x2 = torch.tensor([4, 5, 6])


y1 = torch.stack((x1, x2), dim=1)
z1 = torch.hstack((x1, x2)) #hstack不是叠加而是拼接
#y2 = torch.stack((x1, x2), dim=2) #出错
z2 = torch.dstack((x1,x2)) #dstack是叠加

print(x1.shape, x2.shape)
print(y1,y1.shape)
print(z1,z1.shape)

print(z2,z2.shape)

torch.split(tensor, split_size_or_sections, dim=0)

import torch
x = torch.arange(10).reshape(5,2)

y1 = torch.split(x, 2)# 将dim0分成2
y2 = torch.split(x, [1,4])# 将dim0一四分

print(y1)
print(y2)

torch.squeeze(input, dim=None) → Tensor

import torch
x = torch.zeros(2, 1, 2, 1, 2)

y1 = torch.squeeze(x)
y2 = torch.squeeze(x,0)
y3 = torch.squeeze(x,(1,2,3))

print(x.shape)
print(y1.shape)
print(y2.shape)
print(y3.shape)