Files
pytorch/docs/source/cudnn_persistent_rnn.rst
Rodrigo Berriel 00b732e98b Remove orphan from cuDNN persistent note (#65160)
Summary:
Fixes https://github.com/pytorch/pytorch/issues/60009.

As the document is properly [included](https://github.com/pytorch/pytorch/blob/master/torch/nn/modules/rnn.py#L799), and [`:orphan:` doesn't need to be used in included documents](https://github.com/sphinx-doc/sphinx/issues/6787#issuecomment-549256840), and no warning is emitted in my local build when removing it, I think it can be removed.

The artifact reported in https://github.com/pytorch/pytorch/issues/60009 can be seen in 3 pages: [torch.nn.RNN](https://pytorch.org/docs/stable/generated/torch.nn.RNN.html#torch.nn.RNN), [torch.nn.LSTM](https://pytorch.org/docs/stable/generated/torch.nn.LSTM.html#torch.nn.LSTM), and [torch.nn.GRU](https://pytorch.org/docs/stable/generated/torch.nn.GRU.html#torch.nn.GRU).

cc ezyang suo

Pull Request resolved: https://github.com/pytorch/pytorch/pull/65160

Reviewed By: bdhirsh

Differential Revision: D31020280

Pulled By: ezyang

fbshipit-source-id: 6c3541e5a856a91cf1ce1d2db4d04f5d13118ee4
2021-09-21 11:09:47 -07:00

10 lines
305 B
ReStructuredText

.. note::
If the following conditions are satisfied:
1) cudnn is enabled,
2) input data is on the GPU
3) input data has dtype ``torch.float16``
4) V100 GPU is used,
5) input data is not in ``PackedSequence`` format
persistent algorithm can be selected to improve performance.