PyTorch 张量Basicsoperation

import torch

# creation标量
s = torch.tensor(5)
print("标量:", s)
print("标量维度:", s.ndim)
print("标量形状:", s.shape)

# creation向量
v = torch.tensor([1, 2, 3])
print("\n向量:", v)
print("向量维度:", v.ndim)
print("向量形状:", v.shape)

# creation矩阵
m = torch.tensor([[1, 2], [3, 4]])
print("\n矩阵:", m)
print("矩阵维度:", m.ndim)
print("矩阵形状:", m.shape)

# creation3维张量
t = torch.tensor([[[1, 2], [3, 4]], [[5, 6], [7, 8]]])
print("\n3维张量:", t)
print("3维张量维度:", t.ndim)
print("3维张量形状:", t.shape)

import numpy as np
import torch

# creationNumPyarray
np_array = np.array([[1, 2], [3, 4]])
print("NumPyarray:", np_array)

#  from NumPyarraycreation张量
tensor_from_np = torch.from_numpy(np_array)
print("\n from NumPycreation 张量:", tensor_from_np)
print("张量形状:", tensor_from_np.shape)

#  from 张量creationNumPyarray
np_array_from_tensor = tensor_from_np.numpy()
print("\n from 张量creation NumPyarray:", np_array_from_tensor)

import torch

# creation全零张量
zeros = torch.zeros((2, 3))
print("全零张量:", zeros)

# creation全一张量
ones = torch.ones((2, 3))
print("\n全一张量:", ones)

# creation单位矩阵
eye = torch.eye(3)
print("\n单位矩阵:", eye)

# creation随机张量 (均匀分布) 
rand = torch.rand((2, 3))
print("\n均匀分布随机张量:", rand)

# creation随机张量 (正态分布) 
randn = torch.randn((2, 3))
print("\n正态分布随机张量:", randn)

# creation指定范围 张量
arange = torch.arange(0, 10, step=2)
print("\n指定范围张量:", arange)

# creation线性空间张量
linspace = torch.linspace(0, 1, steps=5)
print("\n线性空间张量:", linspace)

import torch

# creation两个张量
a = torch.tensor([1, 2, 3])
b = torch.tensor([4, 5, 6])

# 加法
print("加法:", a + b)
print("加法:", torch.add(a, b))

# 减法
print("\n减法:", a - b)
print("减法:", torch.sub(a, b))

# 乘法 (元素级) 
print("\n乘法:", a * b)
print("乘法:", torch.mul(a, b))

# 除法 (元素级) 
print("\n除法:", a / b)
print("除法:", torch.div(a, b))

# 矩阵乘法
c = torch.tensor([[1, 2], [3, 4]])
d = torch.tensor([[5, 6], [7, 8]])
print("\n矩阵乘法:", torch.matmul(c, d))
print("矩阵乘法:", c @ d)

import torch

# creation一个张量
t = torch.tensor([[1, 2, 3], [4, 5, 6], [7, 8, 9]])
print("原始张量:", t)

# 访问单个元素
print("\n访问单个元素:", t[0, 0])

# 切片operation
print("\n切片operation - 第一行:", t[0, :])
print("切片operation - 第一列:", t[:, 0])
print("切片operation - 子矩阵:", t[0:2, 0:2])

# advancedindex
print("\nadvancedindex:", t[[0, 2], [1, 2]])

# 掩码index
mask = t > 5
print("\n掩码:", mask)
print("掩码index:", t[mask])

import torch

# creation一个张量
t = torch.tensor([[1, 2, 3], [4, 5, 6]])
print("原始张量:", t)
print("原始形状:", t.shape)

# reshape - 改变形状
reshaped = t.reshape(3, 2)
print("\nreshape after :", reshaped)
print("reshape after 形状:", reshaped.shape)

# view - 改变视graph (共享memory) 
viewed = t.view(3, 2)
print("\nview after :", viewed)
print("view after 形状:", viewed.shape)

# squeeze - 移除维度 for 1 维度
squeezed = torch.tensor([[1, 2, 3]]).squeeze()
print("\nsqueeze after :", squeezed)
print("squeeze after 形状:", squeezed.shape)

# unsqueeze - 添加维度 for 1 维度
unsqueezed = squeezed.unsqueeze(0)
print("\nunsqueeze after :", unsqueezed)
print("unsqueeze after 形状:", unsqueezed.shape)

# transpose - 转置
Transposed = t.transpose(0, 1)
print("\ntranspose after :", Transposed)
print("transpose after 形状:", Transposed.shape)

# permute - 排列维度
permuted = t.permute(1, 0)
print("\npermute after :", permuted)
print("permute after 形状:", permuted.shape)

import torch

# creation两个不同形状 张量
a = torch.tensor([[1, 2, 3], [4, 5, 6]])  # 形状 (2, 3)
b = torch.tensor([10, 20, 30])  # 形状 (3,)

# 广播运算
result = a + b
print("广播加法结果:", result)
print("结果形状:", result.shape)

import torch

# creation不同dataclass型 张量
t_float32 = torch.tensor([1, 2, 3], dtype=torch.float32)
print("float32张量:", t_float32)
print("dataclass型:", t_float32.dtype)

t_int64 = torch.tensor([1, 2, 3], dtype=torch.int64)
print("\nint64张量:", t_int64)
print("dataclass型:", t_int64.dtype)

# class型转换
t_converted = t_float32.to(torch.int64)
print("\n转换 after 张量:", t_converted)
print("转换 after dataclass型:", t_converted.dtype)

import torch

# check is 否 has 可用 GPU
print(" is 否 has 可用GPU:", torch.cuda.is_available())
print("GPU数量:", torch.cuda.device_count())

# creationCPU on  张量
t_cpu = torch.tensor([1, 2, 3])
print("\nCPU张量设备:", t_cpu.device)

# such as果 has GPU, 将张量移 to GPU
if torch.cuda.is_available():
    t_gpu = t_cpu.to('cuda')
    print("GPU张量设备:", t_gpu.device)
    
    # 将张量移回CPU
    t_back_to_cpu = t_gpu.to('cpu')
    print("移回CPU 张量设备:", t_back_to_cpu.device)