permute函数

permute():将tensor的维度换位。

permute(多维数组,[维数换位顺序])

例1:

a=rand(2,3,4);  %a是一个三维数组，各维的长度分别为：2，3，4

permute(a,[2,1,3])  %实现了交换第一维和第二维，a变成3*2*4的矩阵 

例2：
来自行人重识别的一段代码：

	def forward(self, x):
		b, c, h, w = x.size()
		
		if self.use_spatial:
			# spatial attention
			#print(x.shape)
			theta_xs = self.theta_spatial(x)  #[8, 256, 64, 32]->[8, 32, 64, 32]	
			#print(theta_xs.shape)
			phi_xs = self.phi_spatial(x)  #[8, 256, 64, 32]->[8, 32, 64, 32]	
			#print(phi_xs.shape)
			theta_xs = theta_xs.view(b, self.inter_channel, -1) # [8, 32, 64, 32]->[8, 32, 64*32]	
			#print(theta_xs.shape)
			theta_xs = theta_xs.permute(0, 2, 1) # [8, 32, 64*32]->[8 64*32 32]    64*32=2048
			#print(theta_xs.shape)
			phi_xs = phi_xs.view(b, self.inter_channel, -1) # 8 32 64*32	
			#print(phi_xs.shape)
			Gs = torch.matmul(theta_xs, phi_xs)# 8 2048 2048 	
			#print(Gs.shape)