torch.normal function usage

Usage given by official documentation:

torch.normal(means, std, out=None)

means the mean range and shape of the given tensor, std gives the standard deviation of each mean

Official example:

torch.normal(means=torch.arange(1, 11), std=torch.arange(1, 0, -0.1))1.51041.69552.48954.91854.98956.91557.36838.18368.71649.8916[torch.FloatTensor of size 10]

Actual situation:

>>> torch.normal(means=torch.arange(1, 11), std=torch.arange(1, 0, -0.1))Traceback (most recent call last):File "", line 1, in TypeError: normal() received an invalid combination of arguments - got (means=Tensor, std=Tensor, ), but expected one of:* (Tensor mean, Tensor std, *, torch.Generator generator, Tensor out)* (Tensor mean, float std, *, torch.Generator generator, Tensor out)* (float mean, Tensor std, *, torch.Generator generator, Tensor out)* (float mean, float std, tuple of ints size, *, torch.Generator generator, Tensor out, torch.dtype dtype, torch.layout layout, torch.device device, bool pin_memory, bool requires_grad)

From the error type, it can be seen that means is no longer used in torch but mean is used, that is, only a normal distribution with a fixed mean can be generated.

The correct way to open:

Use the fourth option:

torch.normal(mean, std, size)

The three parameters are mean, standard deviation and size

>>> torch.normal(3, 0.1, (3, 4))tensor([[2.9425, 3.1877, 2.9735, 3.0982],[3.0061, 2.9918, 2.7953, 3.0066],[2.8219, 2.9578, 2.8813, 2.9014]])

Use the third option:

torch.normal(mean, stds)

The two parameters are: mean and standard deviation, the standard deviation is used to determine the range size

>>> torch.normal(3, torch.ones(3, 4)/10)tensor([[2.8491, 3.0263, 3.0888, 3.0818],[3.1101, 2.7490, 3.1847, 3.0861],[2.8530, 2.8666, 2.9634, 3.1875]])