[Phi] move ops: maxout/take_along_axis/put_along_axis #39959
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! |
m3ngyang force-pushed the mv_xx_axis_op branch from f84aed8 to bb975e5 Compare m3ngyang force-pushed the mv_xx_axis_op branch from bb975e5 to 0556f4e Compare m3ngyang force-pushed the mv_xx_axis_op branch from 0556f4e to 98b829f Compare m3ngyang force-pushed the mv_xx_axis_op branch 3 times, most recently from e6acbcf to a457ba8 Compare m3ngyang force-pushed the mv_xx_axis_op branch 2 times, most recently from 8193eb5 to ec3e04f Compare 
m3ngyang force-pushed the mv_xx_axis_op branch from ec3e04f to 8aef993 Compare 
YuanRisheng reviewed Mar 1, 2022
Contributor
YuanRisheng left a comment
YuanRisheng left a comment There was a problem hiding this comment.
这个op_benchmark的脚本有补充吗，没有的话还需要加一下
| const auto& index_type = | ||
| paddle::framework::TransToProtoVarType(index.dtype()); | ||
| if (x_grad) { | ||
| paddle::framework::TensorCopy(out_grad, dev_ctx.GetPlace(), x_grad); |
Member Author
这个三个 op 的 benchmark 都需要加下 |
Member Author
| benchmark: PaddlePaddle/benchmark#1326 |
YuanRisheng approved these changes Mar 8, 2022
Member Author
|
zyfncg reviewed Mar 8, 2022
| ele = ele > x ? ele : x; | ||
| } | ||
| template <typename DeviceContext, typename T> | ||
| void MaxOutFunctor<DeviceContext, T>::operator()(const DeviceContext& context, |
Contributor
There was a problem hiding this comment.
这个MaxOutFunctor看了下只有maxout_op在用，建议也迁移到phi下
| // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
| // See the License for the specific language governing permissions and | ||
| // limitations under the License. | ||
| |
| // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
| // See the License for the specific language governing permissions and | ||
| // limitations under the License. | ||
| |
| true, | ||
| errors::PreconditionNotMet("PutAlongAxisGradOpKernel only runs on CPU.")); | ||
| | ||
| const auto& index_type = |
Contributor
There was a problem hiding this comment.
这里不需要转成ProtoVarType了，可以直接用DataType类型进行判断
| | ||
| if (value_grad) { | ||
| value_grad->Resize(index.dims()); | ||
| value_grad->mutable_data<T>(dev_ctx.GetPlace()); |
Contributor
There was a problem hiding this comment.
建议使用dev_ctx.template Alloc<T>()
| } | ||
| if (value_grad) { | ||
| value_grad->Resize(index.dims()); | ||
| value_grad->mutable_data<T>(dev_ctx.GetPlace()); |
| errors::PreconditionNotMet( | ||
| "PutAlongAxisCUDAKernel only runs on GPU device.")); | ||
| | ||
| const auto& index_type = |
Comment on lines +19 to +21
| KernelSignature MaxoutArgumentMapping(const ArgumentMappingContext& ctx) { | ||
| return KernelSignature("maxout", {"X"}, {"groups", "axis"}, {"Out"}); | ||
| } |
Contributor
There was a problem hiding this comment.
这里的前向ArgumentMapping感觉可以不写，使用默认的op_proto应该也能work
Comment on lines +19 to +24
| KernelSignature PutAlongAxisArgumentMapping(const ArgumentMappingContext& ctx) { | ||
| return KernelSignature("put_along_axis", | ||
| {"Input", "Index", "Value"}, | ||
| {"Axis", "Reduce"}, | ||
| {"Result"}); | ||
| } |
Comment on lines +19 to +23
| KernelSignature TakeAlongAxisArgumentMapping( | ||
| const ArgumentMappingContext& ctx) { | ||
| return KernelSignature( | ||
| "take_along_axis", {"Input", "Index"}, {"Axis"}, {"Result"}); | ||
| } |
raindrops2sea approved these changes Mar 8, 2022
chenwhql approved these changes Mar 8, 2022
Contributor
chenwhql left a comment
chenwhql left a comment There was a problem hiding this comment.
LGTM for PADDLE_ENFORCE
Contributor
| 还有一些细节问题，麻烦后续PR追加完善 |
Member Author
ok，下个 pr 补充下 |
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
6 participants
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
Others
Describe
move the following ops: