pytorch基本语法学习（二）

pytorch基本语法学习（二）

经过pytorch基本语法学习（一）的学习，已经对tensor有的基本的认识，在这节的学习中，将是非常非常重要的一节，因为在深度学习图像领域是经常使用到的知识。整装待发，进入学习状态…
1. tensor维度变化 view

a = torch.FloatTensor(2,3,4)
print('a = ', a)  #a =  tensor([[[ 1.4013e-45,  0.0000e+00,  0.0000e+00,  0.0000e+00],
                            # [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
                            # [-1.2391e-11,  3.0756e-41,  0.0000e+00,  0.0000e+00]],

                            # [[ 1.4013e-45,  0.0000e+00,  1.6816e-44,  0.0000e+00],
                            # [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
                            # [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00]]])
print(a.shape)  # torch.Size([2, 3, 4])
print(a.view(2,3*4))  #tensor([[ 1.4013e-45,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
                            # 0.0000e+00,  0.0000e+00,  0.0000e+00, -1.2391e-11,  3.0756e-41,
                            # 0.0000e+00,  0.0000e+00],
                            # [ 1.4013e-45,  0.0000e+00,  1.6816e-44,  0.0000e+00,  0.0000e+00,
                            # 0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
                            # 0.0000e+00,  0.0000e+00]])
print(a.view(2,3*4).shape) # torch.Size([2, 12])
print(a.view(2*3,4))  # tensor([[ 1.4013e-45,  0.0000e+00,  0.0000e+00,  0.0000e+00],
                            # [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
                            # [-1.2391e-11,  3.0756e-41,  0.0000e+00,  0.0000e+00],
                            # [ 1.4013e-45,  0.0000e+00,  1.6816e-44,  0.0000e+00],
                            # [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
                            # [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00]])
print(a.view(2*3,4).shape) # torch.Size([6, 4])
print(a.view(2*3*4))  # tensor([ 1.4013e-45,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
                            # 0.0000e+00,  0.0000e+00,  0.0000e+00, -1.2391e-11,  3.0756e-41,
                            # 0.0000e+00,  0.0000e+00,  1.4013e-45,  0.0000e+00,  1.6816e-44,
                            # 0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
                            # 0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00])
print(a.view(2*3*4).shape)  # torch.Size([24])

b = a.view(2*3*4)
print(b.view(2,3,4))  # tensor([[[ 1.4013e-45,  0.0000e+00,  0.0000e+00,  0.0000e+00],
                             # [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
                             # [-1.2391e-11,  3.0756e-41,  0.0000e+00,  0.0000e+00]],

                             # [[ 1.4013e-45,  0.0000e+00,  1.6816e-44,  0.0000e+00],
                             # [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
                             # [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00]]])
print(b.view(2,3,4).shape)  # torch.Size([2, 3, 4])
print(b.view(3,4,2))  # tensor([[[ 1.4013e-45,  0.0000e+00],
                             # [ 0.0000e+00,  0.0000e+00],
                             # [ 0.0000e+00,  0.0000e+00],
                             # [ 0.0000e+00,  0.0000e+00]],

                             # [[-1.2391e-11,  3.0756e-41],
                             # [ 0.0000e+00,  0.0000e+00],
                             # [ 1.4013e-45,  0.0000e+00],
                             # [ 1.6816e-44,  0.0000e+00]],

                            # [[ 0.0000e+00,  0.0000e+00],
                            # [ 0.0000e+00,  0.0000e+00],
                            # [ 0.0000e+00,  0.0000e+00],
                            # [ 0.0000e+00,  0.0000e+00]]])
print(b.view(3,4,2).shape)  # torch.Size([3, 4, 2])

2. tensor维度变化 squeeze&unsqueeze：压缩或扩展维度，注意只能将dim=1的位置压缩，且扩展也是在指定的位置上扩展为dim=1

a = torch.FloatTensor(2, 3, 4)
print(a.shape)  # torch.Size([2, 3, 4])
print(a.unsqueeze(0).shape)  # torch.Size([1, 2, 3, 4])
print(a.unsqueeze(1).shape)  # torch.Size([2, 1, 3, 4])
print(a.unsqueeze(2).shape)  # torch.Size([2, 3, 1, 4])
print(a.unsqueeze(3).shape)  # torch.Size([2, 3, 4, 1])

b = torch.FloatTensor(1, 2, 1, 1)
print(b.shape)  # torch.Size([1, 2, 1, 1])
print(b.squeeze(0).shape)  # torch.Size([2, 1, 1])
print(b.squeeze(2).shape)  # torch.Size([1, 2, 1])
print(b.squeeze(3).shape)  # torch.Size([1, 2, 1])

3. tensor维度变化 expand & repeat：如果没有特殊需求本人建议使用expend，因为repeat会占用额外内存，注意观察expand & repeat输出shape的区别

b = torch.FloatTensor(1, 2, 1, 1)
print(b.shape)  # torch.Size([1, 2, 1, 1])
print(b.expand(4,2,5,6).shape)  # torch.Size([4, 2, 5, 6])
print(b.repeat(4,2,1,1).shape)  # torch.Size([4, 4, 1, 1])
print(b.repeat(4,1,2,2).shape)  # torch.Size([4, 2, 2, 2])

4. tensor维度变化transpose & permute

#注意观察b，c的转换，虽然a，b，c的shape是相同的，但是b与a并不相同，这也是容易出现错误的地方。
a = torch.FloatTensor(4, 3, 32, 32)
print(a.transpose(1,3).shape)  # torch.Size([4, 32, 32, 3])  维度1和3换位置
b = a.transpose(1,3).contiguous().view(4,3*32*32).view(4, 3, 32, 32)
print(b.shape)  # torch.Size([4, 3, 32, 32])
c = a.transpose(1,3).contiguous().view(4,3*32*32).view(4, 32, 32, 3).transpose(1, 3)
print(c.shape)  # torch.Size([4, 3, 32, 32])
print(torch.all(torch.eq(a,b)))  # tensor(False)  判断a，b是否相等
print(torch.all(torch.eq(a,c)))  # tensor(True)   判断a，c是否相等

c = torch.FloatTensor(4, 3, 28, 32)
print(c.transpose(1,3).shape)  # torch.Size([4, 32, 28, 3])
print(c.transpose(1,3).transpose(1,2).shape)  # torch.Size([4, 28, 32, 3])
print(c.permute(0,2,3,1).shape)  # torch.Size([4, 28, 32, 3])
print(torch.all(torch.eq(c.permute(0,2,3,1),c.transpose(1,3).transpose(1,2))))  # tensor(True)

在下一个学习中主要练习一下tensor的拼接与拆分，数学运算和统计。