view、reshape
- 兩者功能一樣：將數據依次展開后，再變形
- 變形后的數據量與變形前數據量必須相等。即滿足維度：ab...f = xy...z
- reshape是pytorch根據numpy中的reshape來的
- -1表示，其他維度數據已給出情況下，
import torch
a = torch.rand(2, 3, 2, 3)
a
# 輸出：
tensor([[[[0.4850, 0.0073, 0.8941],
[0.0208, 0.4396, 0.7841]],
[[0.0553, 0.3554, 0.0726],
[0.9669, 0.3918, 0.9356]],
[[0.6169, 0.2080, 0.1028],
[0.5816, 0.3509, 0.6983]]],
[[[0.7545, 0.8693, 0.4751],
[0.2206, 0.3384, 0.2877]],
[[0.9521, 0.6172, 0.5058],
[0.6835, 0.0624, 0.6261]],
[[0.7752, 0.3820, 0.5585],
[0.1547, 0.1420, 0.6051]]]])
# 將數據依次展開后，再變形為對應維度。所以，數據的順序是一樣的。且總的數據維度是不變的。
a.view(4, 9)
# 輸出：
tensor([[0.4850, 0.0073, 0.8941, 0.0208, 0.4396, 0.7841, 0.0553, 0.3554, 0.0726],
[0.9669, 0.3918, 0.9356, 0.6169, 0.2080, 0.1028, 0.5816, 0.3509, 0.6983],
[0.7545, 0.8693, 0.4751, 0.2206, 0.3384, 0.2877, 0.9521, 0.6172, 0.5058],
[0.6835, 0.0624, 0.6261, 0.7752, 0.3820, 0.5585, 0.1547, 0.1420, 0.6051]])
# reshape和view的效果一樣
a.reshape(4, 9)
# 輸出：
tensor([[0.4850, 0.0073, 0.8941, 0.0208, 0.4396, 0.7841, 0.0553, 0.3554, 0.0726],
[0.9669, 0.3918, 0.9356, 0.6169, 0.2080, 0.1028, 0.5816, 0.3509, 0.6983],
[0.7545, 0.8693, 0.4751, 0.2206, 0.3384, 0.2877, 0.9521, 0.6172, 0.5058],
[0.6835, 0.0624, 0.6261, 0.7752, 0.3820, 0.5585, 0.1547, 0.1420, 0.6051]])
# -1表示，除了其他維度以外的最大維度數據量
# 此處為2*3*2*3=36, 36/4=9
a.reshape(4, -1)
# 輸出：
tensor([[0.4850, 0.0073, 0.8941, 0.0208, 0.4396, 0.7841, 0.0553, 0.3554, 0.0726],
[0.9669, 0.3918, 0.9356, 0.6169, 0.2080, 0.1028, 0.5816, 0.3509, 0.6983],
[0.7545, 0.8693, 0.4751, 0.2206, 0.3384, 0.2877, 0.9521, 0.6172, 0.5058],
[0.6835, 0.0624, 0.6261, 0.7752, 0.3820, 0.5585, 0.1547, 0.1420, 0.6051]])
# 36/(2*3) = 6
# 所以此時-1表示6
a.reshape(2, 3, -1)
# 輸出：
tensor([[[0.4850, 0.0073, 0.8941, 0.0208, 0.4396, 0.7841],
[0.0553, 0.3554, 0.0726, 0.9669, 0.3918, 0.9356],
[0.6169, 0.2080, 0.1028, 0.5816, 0.3509, 0.6983]],
[[0.7545, 0.8693, 0.4751, 0.2206, 0.3384, 0.2877],
[0.9521, 0.6172, 0.5058, 0.6835, 0.0624, 0.6261],
[0.7752, 0.3820, 0.5585, 0.1547, 0.1420, 0.6051]]])
# 給定的-1以外的其他維度數據的乘積，必須能整除總的數據量
# 比如，此處的5不能整除36，所以會報錯
a.reshape(5, -1)
# 輸出：
---------------------------------------------------------------------------
RuntimeError Traceback (most recent call last)
<ipython-input-26-af98f63a60b8> in <module>
----> 1 a.reshape(5, -1)
RuntimeError: shape '[5, -1]' is invalid for input of size 36
# 變成其他形狀后，如果記得原數據形狀，可以變形回去
a.view(4, 9).view(2, 3, 2, 3)
# 輸出：
tensor([[[[0.4850, 0.0073, 0.8941],
[0.0208, 0.4396, 0.7841]],
[[0.0553, 0.3554, 0.0726],
[0.9669, 0.3918, 0.9356]],
[[0.6169, 0.2080, 0.1028],
[0.5816, 0.3509, 0.6983]]],
[[[0.7545, 0.8693, 0.4751],
[0.2206, 0.3384, 0.2877]],
[[0.9521, 0.6172, 0.5058],
[0.6835, 0.0624, 0.6261]],
[[0.7752, 0.3820, 0.5585],
[0.1547, 0.1420, 0.6051]]]])
unsqueeze
- 功能：指定維度，為其增加（插入）1個維度
- 必須給定維度數據，不然會報錯
a = torch.rand(2, 3)
a
# 輸出：
tensor([[0.3074, 0.2152, 0.0082],
[0.2831, 0.9236, 0.2705]])
# 在0維處，增加1個維度。
a.unsqueeze(0)
# 支持負數索引，等價於 a.unsqueeze(-3)
# 輸出：
tensor([[[0.3074, 0.2152, 0.0082],
[0.2831, 0.9236, 0.2705]]])
a.unsqueeze(1)
# 等價於a.unsqueeze(-2)
# 輸出：
tensor([[[0.3074, 0.2152, 0.0082]],
[[0.2831, 0.9236, 0.2705]]])
a.unsqueeze(2)
# 等價於a.unsqueeze(-1)
# 輸出：
tensor([[[0.3074],
[0.2152],
[0.0082]],
[[0.2831],
[0.9236],
[0.2705]]])
a.unsqueeze(0).unsqueeze(-1) # 分別在0和最后維度上面都增加了一個維度，相比上面的數據，外層多了一對[]括號
# 輸出：
tensor([[[[0.3074],
[0.2152],
[0.0082]],
[[0.2831],
[0.9236],
[0.2705]]]])
squeeze 刪除為1的維度
- 默認將刪除所有為1的維度
- 傳入指定維度，刪除指定維度為1的維度
a = torch.rand(2, 1, 3, 1, 2, 1)
a
# 輸出：
tensor([[[[[[0.4492],
[0.6223]]],
[[[0.8522],
[0.4971]]],
[[[0.2585],
[0.6034]]]]],
[[[[[0.1904],
[0.5564]]],
[[[0.3120],
[0.2698]]],
[[[0.6620],
[0.5963]]]]]])
a.squeeze() # 默認刪除所有為1的維度，即(2, 1, 3, 1, 2, 1)刪除后得到（2, 3, 2）
a
# 輸出：
tensor([[[0.4492, 0.6223],
[0.8522, 0.4971],
[0.2585, 0.6034]],
[[0.1904, 0.5564],
[0.3120, 0.2698],
[0.6620, 0.5963]]])
a.squeeze(1).squeeze(2).squeeze(3) # 通過指定維度刪除1，與上面效果相同
a
# 輸出：
tensor([[[0.4492, 0.6223],
[0.8522, 0.4971],
[0.2585, 0.6034]],
[[0.1904, 0.5564],
[0.3120, 0.2698],
[0.6620, 0.5963]]])
expand
- 功能：擴展維度到指定維度。擴展的數據是復制當前維度的數據。
- 只有只有維度大小為1的維度可以擴展
- -1表示保持原維度
a = torch.rand(2, 3, 1)
a
# 輸出：
tensor([[[0.1162],
[0.6026],
[0.5674]],
[[0.7272],
[0.4351],
[0.1708]]])
# 擴展數據，只有維度為1的地方可以被擴展
a.expand(2, -1, 3)
# 輸出：
tensor([[[0.1162, 0.1162, 0.1162],
[0.6026, 0.6026, 0.6026],
[0.5674, 0.5674, 0.5674]],
[[0.7272, 0.7272, 0.7272],
[0.4351, 0.4351, 0.4351],
[0.1708, 0.1708, 0.1708]]])
repeat
- 表示擴展數據到原維度的倍數。
- 1表示保持原維度。
- 不建議使用此方法，因為復制數據會重新申請內存空間，比較占內存
a
# 輸出：
tensor([[[0.1162],
[0.6026],
[0.5674]],
[[0.7272],
[0.4351],
[0.1708]]])
a.repeat(1, 1, 3)
# 輸出：
tensor([[[0.1162, 0.1162, 0.1162],
[0.6026, 0.6026, 0.6026],
[0.5674, 0.5674, 0.5674]],
[[0.7272, 0.7272, 0.7272],
[0.4351, 0.4351, 0.4351],
[0.1708, 0.1708, 0.1708]]])
矩陣轉置
- 行列互換
- .t()只支持2D的數據
a = torch.rand(5, 2)
a
# 輸出：
tensor([[0.9841, 0.2180],
[0.5082, 0.8553],
[0.5250, 0.7228],
[0.9064, 0.0074],
[0.2752, 0.3939]])
a.t()
# 輸出：
tensor([[0.9841, 0.5082, 0.5250, 0.9064, 0.2752],
[0.2180, 0.8553, 0.7228, 0.0074, 0.3939]])
transpose維度交換
- 相當於被選擇的兩個維度數據進行轉置，即：行列互換
- 也可以理解為，每一行數據順時針旋轉90度為一列后，依次拼接在右邊（最后面）
a = torch.rand(2, 3, 4)
a
# 輸出：
tensor([[[0.1696, 0.6733, 0.7269, 0.1066],
[0.5433, 0.2820, 0.0214, 0.9471],
[0.6922, 0.8220, 0.8422, 0.8682]],
[[0.6121, 0.8133, 0.6502, 0.4529],
[0.9810, 0.9233, 0.5279, 0.2193],
[0.1775, 0.8487, 0.4938, 0.3994]]])
a.transpose(0, 2)
# 輸出：
tensor([[[0.1696, 0.6121],
[0.5433, 0.9810],
[0.6922, 0.1775]],
[[0.6733, 0.8133],
[0.2820, 0.9233],
[0.8220, 0.8487]],
[[0.7269, 0.6502],
[0.0214, 0.5279],
[0.8422, 0.4938]],
[[0.1066, 0.4529],
[0.9471, 0.2193],
[0.8682, 0.3994]]])
a.transpose(0, 2).contiguous().view(4, 6)
# 輸出：
tensor([[0.1696, 0.6121, 0.5433, 0.9810, 0.6922, 0.1775],
[0.6733, 0.8133, 0.2820, 0.9233, 0.8220, 0.8487],
[0.7269, 0.6502, 0.0214, 0.5279, 0.8422, 0.4938],
[0.1066, 0.4529, 0.9471, 0.2193, 0.8682, 0.3994]])
permute
與transpose不同的是：
- transpose一次只能進行兩個維度之間的交換
- permute可以一次進行多個維度的交換。其實質就是實現多個transpose步驟。
- 給定維度索引。
a = torch.rand(2, 3, 4)
a
# 輸出：
tensor([[[0.7855, 0.6239, 0.3785, 0.8138],
[0.9595, 0.5210, 0.3816, 0.1612],
[0.0152, 0.9714, 0.4245, 0.4754]],
[[0.1475, 0.4961, 0.0812, 0.3769],
[0.4279, 0.0595, 0.0717, 0.9871],
[0.9480, 0.3525, 0.3076, 0.0367]]])
a.permute(1, 2, 0)
# 輸出：
tensor([[[0.7855, 0.1475],
[0.6239, 0.4961],
[0.3785, 0.0812],
[0.8138, 0.3769]],
[[0.9595, 0.4279],
[0.5210, 0.0595],
[0.3816, 0.0717],
[0.1612, 0.9871]],
[[0.0152, 0.9480],
[0.9714, 0.3525],
[0.4245, 0.3076],
[0.4754, 0.0367]]])
# 通過2次transpose可以實現permute的效果
a.transpose(0, 1).transpose(1, 2)
# 輸出：
tensor([[[0.7855, 0.1475],
[0.6239, 0.4961],
[0.3785, 0.0812],
[0.8138, 0.3769]],
[[0.9595, 0.4279],
[0.5210, 0.0595],
[0.3816, 0.0717],
[0.1612, 0.9871]],
[[0.0152, 0.9480],
[0.9714, 0.3525],
[0.4245, 0.3076],
[0.4754, 0.0367]]])
broadcast廣播（自動擴展）：
- 定義：兩個維度不同的數據相加，會自動將維度小的數據擴張到與大維度相同的維度
- 特點：自動擴展維度，並且不用復制數據，節省空間
- 前提：小維度的數據，要么不給定維度，要么給定1維，要么與原數據維度大小相同的維度，否則會報錯。
a = torch.rand(3, 4, 2)
a
# 輸出：
tensor([[[0.7220, 0.7834],
[0.4305, 0.6128],
[0.6032, 0.5751],
[0.2304, 0.3003]],
[[0.1044, 0.1071],
[0.1373, 0.4874],
[0.3963, 0.5231],
[0.1851, 0.8962]],
[[0.0677, 0.0587],
[0.7268, 0.8807],
[0.5445, 0.2110],
[0.8755, 0.8577]]])
b = torch.rand(1, 4, 2) # b的第一維度為1，a的為3
b
# 輸出：
tensor([[[0.6676, 0.5702],
[0.3334, 0.8553],
[0.5392, 0.0754],
[0.9488, 0.7814]]])
# 相加和，b被自動廣播為3，然后再與a相加，得到結果
a + b
# 輸出：
tensor([[[1.0925, 1.1274],
[0.7881, 1.6091],
[0.6568, 0.8419],
[1.8460, 1.7360]],
[[0.9783, 0.9697],
[1.0127, 1.3521],
[0.7147, 0.3189],
[1.3435, 1.3845]],
[[0.8699, 1.0997],
[0.5712, 1.8381],
[0.6607, 1.0689],
[0.9599, 1.7686]]])
