frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-10-20 21:14:14 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix typo under aten/src/ATen/cudnn (#96063)

Browse Source 
This files fixes typo in comments under `aten/src/ATen/cudnn` directory
Pull Request resolved: https://github.com/pytorch/pytorch/pull/96063
Approved by: https://github.com/ngimel
This commit is contained in:
Kazuaki Ishizaki
2023-03-07 03:17:29 +00:00
committed by PyTorch MergeBot
parent 19ee61f7fa
commit e9ca902894
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
aten/src/ATen/cudnn/Descriptors.h
View File

@ -135,7 +135,7 @@ class TORCH_CUDA_CPP_API TensorDescriptor : public Descriptor<
// padding). If 't' is lower-dimensional than 'pad', the remaining
// dimensions (on the right) are padded with ones. This doesn't
// affect the underlying data layout. This is particularly useful for
// dealing with a pecularity of the CuDNN API, which is that broadcasting in CuDNN is
// dealing with a peculiarity of the CuDNN API, which is that broadcasting in CuDNN is
// done in two steps: first, the client code is expected to pad out
// (the dimensions) input tensors to be the same dimension as the
// target broadcast, and then second, CuDNN takes of actually
@ -245,7 +245,7 @@ struct TORCH_CUDA_CPP_API DropoutDescriptor
// NB: seed doesn't matter when dropout = 0, because no random number
// initialization actually takes place when there is no dropout.
// NB: Empirically, cudnnSetDropoutDescriptor is cheap when
// dropoot == 0
// dropout == 0
AT_CUDNN_CHECK(cudnnSetDropoutDescriptor(mut_desc(), handle, 0 /* dropout */, nullptr, 0 /* state_size */, 0 /* seed */));
}
};