torch.Tensor.nan_to_num_

Tensor.nan_to_num_(nan=0.0, posinf=None, neginf=None) → Tensor

In-place version of nan_to_num().