pytorch的python API略读--tensor(四）

作者：机器视觉全栈er
来源：cvtutorials.com

2.1.3 分

分就是将tensor拆分或者只看tensor的某个部分。

torch.chunk：将一个tensor分解成多个块。用法如下：

torch.chunk(input, chunks, dim=0)

如果chunk不能被输入的tensor的dim方向上的整除的话，最后一个块和其他块的大小不一样，举个例子来说明下：

>>> cvtutorials = torch.randn(7, 7)
>>> cvtutorials
tensor([[-1.1886, -1.4592, -0.2017, -1.7050,  2.0357, -0.3754, -0.6668],
        [ 0.5493, -0.8917, -0.3982, -2.4488,  1.8564, -0.6771,  0.5145],
        [-0.9599,  0.7301, -1.1945, -0.4188, -0.4440,  1.5929, -0.8686],
        [ 0.0075,  0.3351,  0.2337,  0.6663,  0.7913,  0.0717,  1.0995],
        [ 0.6604,  0.6436,  0.2717, -1.0651,  0.6586, -1.4068, -1.1303],
        [ 0.3116,  0.7408, -0.3726,  1.2334, -0.6076, -1.3587, -1.8686],
        [-0.8115, -1.4766,  1.4367,  2.1566, -1.1522, -0.8364, -1.0398]])
>>> torch.chunk(cvtutorials, 3, 0)
(tensor([[-1.1886, -1.4592, -0.2017, -1.7050,  2.0357, -0.3754, -0.6668],
        [ 0.5493, -0.8917, -0.3982, -2.4488,  1.8564, -0.6771,  0.5145],
        [-0.9599