view関数

以下に使用例を示す.

import torch

# 3x2x2のランダムテンソルを作成
r3 = torch.randn((3, 2, 2))
print("元のテンソル形状:", r3.shape)
print("元のテンソル形状:", r3)    # torch.Size([3, 2, 2])

# viewを使って形状を変更する例
r3_reshaped = r3.view(12)  # 1次元に平坦化
print("1次元化:", r3_reshaped.shape)
print("1次元化:", r3_reshaped)  # torch.Size([12])

r3_reshaped2 = r3.view(3, 4)  # 3x4に変形
print("3x4に変形:", r3_reshaped2.shape)

print("3x4に変形:", r3_reshaped2)    # torch.Size([3, 4])

r3_reshaped3 = r3.view(-1, 2)  # サイズを自動計算（6x2）
print("自動計算で変形:", r3_reshaped3.shape)
print("自動計算で変形:", r3_reshaped3)    # torch.Size([6, 2])

以上により,

元のテンソル形状: tensor([[[ 0.5661, -0.3820],
         [ 0.8807,  0.2710]],

        [[ 0.7694,  0.3453],
         [ 1.8979, -0.2357]],

        [[ 0.7885,  0.3208],
         [ 0.8456, -0.3621]]])
1次元化: tensor([ 0.5661, -0.3820,  0.8807,  0.2710,  0.7694,  0.3453,  1.8979, -0.2357,
         0.7885,  0.3208,  0.8456, -0.3621])
3x4に変形: tensor([[ 0.5661, -0.3820,  0.8807,  0.2710],
        [ 0.7694,  0.3453,  1.8979, -0.2357],
        [ 0.7885,  0.3208,  0.8456, -0.3621]])
自動計算で変形: tensor([[ 0.5661, -0.3820],
        [ 0.8807,  0.2710],
        [ 0.7694,  0.3453],
        [ 1.8979, -0.2357],
        [ 0.7885,  0.3208],
        [ 0.8456, -0.3621]])

Mathematics is the language with which God has written the universe.