Function trunc_normal_

utils.py:528–530

(tensor, mean=0., std=1., a=-2., b=2.)

def trunc_normal_(tensor, mean=0., std=1., a=-2., b=2.):
    # type: (Tensor, float, float, float, float) -> Tensor
    return _no_grad_trunc_normal_(tensor, mean, std, a, b)


class LARS(torch.optim.Optimizer):