e.g. torch.Tensor(3,4), I will get a matrix with shape of 3,4, and I notice that the mean of this matrix is 0, but the what is the specific initialization of this operation?

I would recommend to stick to the factory methods, e.g. `torch.randn`

, `torch.rand`

, `torch.tensor([1, 2, 3])`

etc.

`torch.Tensor`

will return an uninitialized `FloatTensor`

, which might be dangerous.

1 Like