add new api paddle.Tensor.fill_diagonal_ #34460
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
|
Thanks for your contribution! |
jeff41404
reviewed
Aug 2, 2021
Comment on lines
47
to
49
| "offset of diagonal, zero means no offset, positive means " | ||
| "offset to up-right corner; negtive means offset to " | ||
| "bottom-left corner") |
jeff41404
reviewed
Aug 2, 2021
| )DOC"); | ||
| AddInput("X", "(Tensor) The input tensor."); | ||
| AddOutput("Out", | ||
| "Tensor, the clipped tensor, with the same shape and data type " |
There was a problem hiding this comment.
“the clipped tensor”的说明在这里有什么特殊意义不，是否去掉？
jeff41404
reviewed
Aug 6, 2021
Comment on lines
95
to
98
| framework::TensorCopy(*xin, ctx.GetPlace(), out); | ||
|
|
||
| T *out_data = out->mutable_data<T>(ctx.GetPlace()); |
There was a problem hiding this comment.
常规做法是先 out->mutable_data 分配内存后再使用 TensorCopy，这里反过来是否有特殊考虑
There was a problem hiding this comment.
done, TensorCopy中也有out->mutable_data，所以其实都可以。不过也已经修改。
jeff41404
reviewed
Aug 6, 2021
Comment on lines
+61
to
+49
| framework::TensorCopy(*xin, ctx.GetPlace(), out); | ||
|
|
||
| T* out_data = out->mutable_data<T>(ctx.GetPlace()); |
There was a problem hiding this comment.
常规做法是先 out->mutable_data 分配内存后再使用 TensorCopy，这里反过来是否有特殊考虑
jeff41404
reviewed
Aug 6, 2021
| if ((i - offset) % strides == 0 && i < wrapsize) { | ||
| data[i] = T(0); | ||
| } else { | ||
| data[i] = static_cast<T>(doutdata[i]); |
There was a problem hiding this comment.
反向的逻辑中是否也可以使用TensorCopy 批量实现数据cp
jeff41404
reviewed
Aug 6, 2021
| if ((idx - offset) % strides == 0 && idx < wrapsize) { | ||
| dst_data[idx] = T(0); | ||
| } else { | ||
| dst_data[idx] = const_cast<T&>(src_data[idx]); |
There was a problem hiding this comment.
反向的逻辑中是否也可以使用TensorCopy 批量实现数据cp
jiangjiajun
requested changes
Aug 9, 2021
| AddOutput("Out", | ||
| "Tensor, the output tensor, with the same shape and data type " | ||
| "as input(x)"); | ||
| AddAttr<float>( |
There was a problem hiding this comment.
以float类型读取python数值，然后填入时转化为实际类型填入。
| protected: | ||
| void Apply(GradOpPtr<T> retv) const override { | ||
| retv->SetType("fill_diagonal_grad"); | ||
| // retv->SetInput("Out", this->Output("Out")); |
| expected_grad = np.array( | ||
| [[0, 1, 1], [1, 0, 1], [1, 1, 0]]).astype('float32') | ||
|
|
||
| typelist = ['float32', 'float64', 'int32', 'int64'] |
There was a problem hiding this comment.
Kernel注册了Bool类型，也加上bool类型的测试
XiaoguangHu01
previously approved these changes
Aug 13, 2021
| x(Tensor): ``x`` is the original Tensor | ||
| value(Scale): ``value`` is the value to filled in x | ||
| offset(int,optional): the offset to the main diagonal. Default: 0 (main diagonal). | ||
| name(str,optional): Name for the operation (optional, default is None) |
zhiboniu
dismissed stale reviews from XiaoguangHu01, jzhang533, and jiangjiajun
via
a7376d0
zhiboniu
changed the title
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
APIs
Describe
补充缺失API，新增 api paddle.Tensor.fill_diagonal_
功能：实现对Tensor对角线向量值的修改。
使用示例：
x = paddle.ones((3,3))
x.fill_diagonal_(5)
tensor([[5., 1., 1.],
[1., 5., 1.],
[1., 1., 5.]])