[pnorm] fix bug in fp16 & optimize memory #39011
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Thanks for your contribution! |
ZHUI
reviewed
Jan 21, 2022
| struct UnsignedPowFunctor { | ||
| HOSTDEVICE explicit inline UnsignedPowFunctor(float porder) { | ||
| this->porder = porder; | ||
| } | ||
| HOSTDEVICE inline Ty operator()(const Tx x) const { |
There was a problem hiding this comment.
原来这里两个类型是要支持 fp16 场景吗？现在为什么不用了
There was a problem hiding this comment.
之前的实现是有问题的，fp16类型在inline function里边会转换为float类型计算，并不需要在外部转换。
| TensorReduceFunctorImpl<T, T, kps::AddFunctor, UnsignedPowFunctor<T>>( | ||
| *in_x, out_norm, UnsignedPowFunctor<T>(porder), reduce_axis, stream); | ||
|
|
||
| const framework::Tensor* tmp_norm = out_norm; |
There was a problem hiding this comment.
这里的tmp_norm不需要了吧，直接是 out_norm
There was a problem hiding this comment.
这里ins的类型是 std::vector<const framework::Tensor*> ，需要转换为const。
paddle/fluid/operators/p_norm_op.cu
Outdated
| dx->device(place) = dy->broadcast(dim) * equals.select(ones, zeros) * | ||
| positives.select(ones, negs); | ||
| dx->device(place) = dy->broadcast(dim) * (*x).sign() * | ||
| ((*x).abs() == y->broadcast(dim)).select(ones, zeros); |
There was a problem hiding this comment.
dx.device(*place) =
(x.abs() == norm.broadcast(bcast)).template cast<T>() * x.sign() *
norm_dy.broadcast(bcast);也可以用cast
There was a problem hiding this comment.
这里意思是按照这种写法cast就可以获取非零值吗
paddle/fluid/operators/p_norm_op.cu
Outdated
| template <typename DeviceContext, typename X, typename Y, typename DX, | ||
| typename DY, typename Dim> | ||
| void operator()(const DeviceContext& place, X* x, Y* y, DX* dx, DY* dy, | ||
| const Dim& dim, int size) { | ||
| auto ones = dx->constant(static_cast<T>(1.)); | ||
| auto negs = dx->constant(static_cast<T>(-1.)); | ||
| auto zeros = dx->constant(static_cast<T>(0.)); |
There was a problem hiding this comment.
auto ones = dx->constant(static_cast<T>(1.));
auto negs = dx->constant(static_cast<T>(-1.));
auto zeros = dx->constant(static_cast<T>(0.));都不需要了吧
wanghuancoder
approved these changes
Jan 25, 2022
Xreki
approved these changes
Jan 25, 2022
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
PR types
Bug fixes
PR changes
OPs
Describe