[Phi] Move QR to Phi #44742
Merged
[Phi] Move QR to Phi #44742
Conversation
This file contains 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
zyfncg
reviewed
Jul 30, 2022
paddle/fluid/operators/lstsq_op.cu
Outdated
| @@ -29,6 +28,29 @@ namespace operators { | |||
|
|
|||
| using paddle::framework::Tensor; | |||
|
|
|||
| template <typename DeviceContext, typename T> | |||
| param : [x] | ||
| kernel : | ||
| func : qr_grad | ||
| # optional: q_grad, r_grad |
paddle/phi/kernels/gpu/qr_kernel.cu
Outdated
| } | ||
|
|
||
| template <typename T, typename Context> | ||
| static DenseTensor TrilTriu(const Context& ctx, |
There was a problem hiding this comment.
这个TrilTriu可以放到 paddle/phi/kernels/tril_triu_kernel.h头文件中，方便其他Kernel复用
paddle/phi/kernels/gpu/qr_kernel.cu
Outdated
Comment on lines
212
to
218
| DenseTensor* workspace = new DenseTensor(); | ||
| workspace->Resize(make_ddim({lwork})); | ||
| float* workspace_ptr = dev_ctx.template Alloc<float>(workspace); | ||
|
|
||
| DenseTensor* info = new DenseTensor(); | ||
| info->Resize(make_ddim({1})); | ||
| int* info_d = dev_ctx.template Alloc<int>(info); |
There was a problem hiding this comment.
workspace和info用指针容易内存泄露，建议使用栈上对象或者unique_ptr替换
paddle/phi/kernels/gpu/qr_kernel.cu
Outdated
Comment on lines
266
to
270
| DenseTensor* workspace = new DenseTensor(); | ||
| workspace->Resize(make_ddim({lwork})); | ||
| double* workspace_ptr = dev_ctx.template Alloc<double>(workspace); | ||
|
|
||
| DenseTensor* info = new DenseTensor(); |
paddle/phi/kernels/gpu/qr_kernel.cu
Outdated
Comment on lines
321
to
327
| DenseTensor* workspace = new DenseTensor(); | ||
| workspace->Resize(make_ddim({lwork})); | ||
| float* workspace_ptr = dev_ctx.template Alloc<float>(workspace); | ||
|
|
||
| DenseTensor* info = new DenseTensor(); | ||
| info->Resize(make_ddim({1})); | ||
| int* info_d = dev_ctx.template Alloc<int>(info); |
paddle/phi/kernels/gpu/qr_kernel.cu
Outdated
Comment on lines
377
to
383
| DenseTensor* workspace = new DenseTensor(); | ||
| workspace->Resize(make_ddim({lwork})); | ||
| double* workspace_ptr = dev_ctx.template Alloc<double>(workspace); | ||
|
|
||
| DenseTensor* info = new DenseTensor(); | ||
| info->Resize(make_ddim({1})); | ||
| int* info_d = dev_ctx.template Alloc<int>(info); |
Comment on lines
51
to
76
| template <typename T, typename Context> | ||
| static DenseTensor TrilTriu(const Context& ctx, | ||
| const DenseTensor& x, | ||
| int diagonal, | ||
| bool lower) { | ||
| DenseTensor dense_out; | ||
| MetaTensor meta_out(&dense_out); | ||
| TrilTriuInferMeta(x, diagonal, lower, &meta_out); | ||
| TrilTriuKernel<T, Context>(ctx, x, diagonal, lower, &dense_out); | ||
| return dense_out; | ||
| } | ||
|
|
||
| template <typename T, typename Context> | ||
| static DenseTensor TriangularSolve(const Context& ctx, | ||
| const DenseTensor& x, | ||
| const DenseTensor& y, | ||
| bool upper, | ||
| bool transpose, | ||
| bool unitriangular) { | ||
| DenseTensor dense_out; | ||
| MetaTensor meta_out(&dense_out); | ||
| TriangularSolveInferMeta(x, y, upper, transpose, unitriangular, &meta_out); | ||
| TriangularSolveKernel<T, Context>( | ||
| ctx, x, y, upper, transpose, unitriangular, &dense_out); | ||
| return dense_out; | ||
| } |
YuanRisheng
reviewed
Aug 1, 2022
paddle/phi/kernels/gpu/qr_kernel.cu
Outdated
| #include <vector> | ||
|
|
||
| #include "paddle/fluid/memory/memcpy.h" | ||
| #include "paddle/fluid/platform/enforce.h" |
YuanRisheng
approved these changes
Aug 2, 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
Others
PR changes
Ops
Describe
This pr moves the QR kernels to Phi library.