Add C++ EinsumOp which support 2 operands einsum. #42105
Merged
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
…ast 3. MultiVariable
Aurelius84
reviewed
Apr 24, 2022
| #include "paddle/fluid/framework/op_registry.h" | ||
| #include "paddle/fluid/framework/operator.h" | ||
| #include "paddle/phi/core/ddim.h" | ||
| #include "paddle/phi/kernels/impl/einsum_impl.h" |
There was a problem hiding this comment.
<memory>，<unordered_map>，， einsum_impl.h这4个头文件是必需的么？
There was a problem hiding this comment.
目前einsum impl中有InferMeta，后续会提PR将这个函数移动到 unary.cc 中，下个删除
paddle/fluid/operators/einsum_op.cc
Outdated
| void Make() override { | ||
| AddInput("Operands", "(Tensor), The input tensor of svd op.") | ||
| .AsDuplicable(); | ||
| AddOutput("Out", "(Tensor), The output VH tensor of svd op."); |
There was a problem hiding this comment.
Suggested change
| AddOutput("Out", "(Tensor), The output VH tensor of svd op."); | |
| AddOutput("Out", "(Tensor), The output tensor of einsum op."); |
paddle/fluid/operators/einsum_op.cc
Outdated
| using framework::OperatorWithKernel::OperatorWithKernel; | ||
| void InferShape(framework::InferShapeContext* ctx) const override { | ||
| auto in_x = "Operands"; | ||
| auto out_x_g_n = framework::GradVarName(in_x); |
There was a problem hiding this comment.
Suggested change
| auto out_x_g_n = framework::GradVarName(in_x); | |
| auto x_name = "Operands"; | |
| auto x_grad_name = framework::GradVarName(x_name); |
注意命名规范
|
|
||
| #include "paddle/phi/backends/gpu/gpu_context.h" | ||
| #include "paddle/phi/core/kernel_registry.h" | ||
| #include "paddle/phi/kernels/einsum_kernel.h" |
There was a problem hiding this comment.
#include "paddle/phi/kernels/einsum_kernel.h" 要放在第一行，且空行隔开
|
|
||
| #include "paddle/phi/backends/gpu/gpu_context.h" | ||
| #include "paddle/phi/core/kernel_registry.h" | ||
| #include "paddle/phi/kernels/einsum_kernel.h" |
There was a problem hiding this comment.
#include "paddle/phi/kernels/einsum_kernel.h" 要放在第一行，且空行隔开
| } else { | ||
| PADDLE_ENFORCE_EQ((*labelshape)[c], | ||
| op_dim[dim_ptr], | ||
| phi::errors::InvalidArgument("")); |
| PADDLE_ENFORCE_EQ(v == 1 || broadcast_dims->at(idx) == 1 || | ||
| broadcast_dims->at(idx) == v, | ||
| true, | ||
| phi::errors::InvalidArgument("can't broad cast.")); |
| const LabelMap& labelshape, | ||
| std::vector<int>* output_dims) { | ||
| for (int c : right) { | ||
| if (c == '.') |
| std::vector<int>* output_dims, | ||
| std::string* right) { | ||
| auto results = paddle::string::split_string(equation, "->"); | ||
| auto left = results[0]; |
There was a problem hiding this comment.
这里是否需要先ENFORCE下 results的size是否符合预期？
There was a problem hiding this comment.
这个步骤在 ValidEinsum 函数中已经验证了。这里就没验证了。
| VLOG(5) << "Einsum Infershape: output dims:" | ||
| << paddle::string::join_strings(output_dims, ","); | ||
| print_label(all_labels, labeltype); | ||
| print_label(all_labels, labelshape); |
There was a problem hiding this comment.
这两个print放到VLOG下来触发吧，另外，op的vlog等级可以设置为3
Aurelius84
reviewed
Apr 24, 2022
| namespace ops = paddle::operators; | ||
|
|
||
| DECLARE_INFER_SHAPE_FUNCTOR(einsum, EinsumInferShapeFunctor, | ||
| PD_INFER_META(phi::EinsumInferShape)); |
XieYunshen
approved these changes
Apr 25, 2022
XieYunshen
approved these changes
Apr 25, 2022
TCChenlong
approved these changes
Apr 25, 2022
chenwhql
approved these changes
Apr 25, 2022
| } | ||
| } | ||
|
|
||
| inline void EinsumInferShape(const std::vector<const MetaTensor*>& inputs, |
chenwhql
reviewed
Apr 25, 2022
| auto it = std::find(all_labels.begin(), all_labels.end(), c); | ||
| PADDLE_ENFORCE_NE(it, | ||
| all_labels.end(), | ||
| phi::errors::InvalidArgument("Must in all_labels.")); |
Aurelius84
approved these changes
Apr 26, 2022
2742195759
added a commit
to 2742195759/Paddle
that referenced
this pull request
Apr 28, 2022
* full api fix * when out is None, go old dygraph mode * by static check * first version: support 2-inputs forwards. TODO: 1. backward 2. BroadCast 3. MultiVariable * time out -> 120
XiaoguangHu01
pushed a commit
that referenced
this pull request
Apr 28, 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
New features
PR changes
APIs
Describe
This PR developed C++ EinsumOp.
TODO: