mirror of
https://github.com/zebrajr/pytorch.git
synced 2025-12-06 12:20:52 +01:00
23b98202b5
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/67721
This moves the operator calls (`expand_inplace` and `to(kCUDA)`) into
`bernoulli_impl_` which is shared between CPU and CUDA. So that the
cuda code only needs to generates random numbers and nothing else.
The other changes are just rearranging includes to avoid including
`Tensor.h`.
Test Plan: Imported from OSS
Reviewed By: mruberry
Differential Revision: D32500426
Pulled By: malfet
fbshipit-source-id: f855c2055392355e79e3df832fa56d2041ebf24e
(cherry picked from commit 7ab7d17e94)
24 lines
883 B
C++
24 lines
883 B
C++
#include <ATen/core/TensorBase.h>
|
|
|
|
// Broadcasting utilities for working with TensorBase
|
|
namespace at {
|
|
namespace internal {
|
|
TORCH_API TensorBase expand_slow_path(const TensorBase &self, IntArrayRef size);
|
|
} // namespace internal
|
|
|
|
inline c10::MaybeOwned<TensorBase> expand_size(const TensorBase &self, IntArrayRef size) {
|
|
if (size.equals(self.sizes())) {
|
|
return c10::MaybeOwned<TensorBase>::borrowed(self);
|
|
}
|
|
return c10::MaybeOwned<TensorBase>::owned(
|
|
at::internal::expand_slow_path(self, size));
|
|
}
|
|
c10::MaybeOwned<TensorBase> expand_size(TensorBase &&self, IntArrayRef size) = delete;
|
|
|
|
inline c10::MaybeOwned<TensorBase> expand_inplace(const TensorBase &tensor, const TensorBase &to_expand) {
|
|
return expand_size(to_expand, tensor.sizes());
|
|
}
|
|
c10::MaybeOwned<TensorBase> expand_inplace(const TensorBase &tensor, TensorBase &&to_expand) = delete;
|
|
|
|
} // namespace at
|