tensorplay.nn.init.trunc_normal_

tensorplay.nn.init.trunc_normal_(tensor, mean=0.0, std=1.0, a=-2.0, b=2.0)[source]

Fills the input Tensor with values drawn from a truncated normal distribution.

Method is based on the rejection-sampling scheme in https://people.sc.fsu.edu/~jburkardt/presentations/truncated_normal.pdf — a direct port of torch.nn.init._no_grad_trunc_normal_ (torch/nn/init.py), which torchvision transformer models use for positional embeddings.