Matthew Hoffman e40d6ae0a7 Improve torch.cuda.amp type hints (#108630) ... Fixes #108629 1. Add the following to their modules' `__all__` so that pyright considers them to be publicly exported: * [`torch.autocast`](https://pytorch.org/docs/stable/amp.html#torch.autocast) * [`torch.cuda.amp.GradScaler`](https://pytorch.org/docs/stable/amp.html#torch.cuda.amp.GradScaler) * [`torch.cuda.amp.autocast`](https://pytorch.org/docs/stable/amp.html#torch.cuda.amp.autocast) * [`torch.cuda.amp.custom_fwd`](https://pytorch.org/docs/stable/amp.html#torch.cuda.amp.custom_fwd) * [`torch.cuda.amp.custom_bwd`](https://pytorch.org/docs/stable/amp.html#torch.cuda.amp.custom_bwd) 2. Add `overload`s for `torch.cuda.amp.GradScaler.scale` to differentiate when a `torch.Tensor` is returned vs. an `Iterable[torch.Tensor]` is returned based on the type of the `outputs` parameter. Pull Request resolved: https://github.com/pytorch/pytorch/pull/108630 Approved by: https://github.com/ezyang		2023-09-08 06:06:25 +00:00
..
__init__.py	Improve torch.cuda.amp type hints (#108630)	2023-09-08 06:06:25 +00:00
autocast_mode.py	Apply UFMT to low traffic torch modules (#106249)	2023-07-29 23:37:30 +00:00
common.py	Apply UFMT to low traffic torch modules (#106249)	2023-07-29 23:37:30 +00:00
grad_scaler.py	Improve torch.cuda.amp type hints (#108630)	2023-09-08 06:06:25 +00:00