映射文档 No.48 #5840

YiZoey · 2023-04-25T05:57:41Z

分组 ID 48

paddle-bot · 2023-04-25T05:57:45Z

感谢你贡献飞桨文档，文档预览构建中，Docs-New 跑完后即可预览，预览链接：http://preview-pr-5840.paddle-docs-preview.paddlepaddle.org.cn/documentation/docs/zh/api/index_cn.html
预览工具的更多说明，请参考：飞桨文档预览工具

YiZoey · 2023-04-25T08:17:33Z

映射任务列表48中的torch.matmul已经被其他贡献者提交；
https://www.paddlepaddle.org.cn/documentation/docs/zh/api/paddle/linalg/inv_cn.html 下代码示例的的函数名称与文档函数不对应；

huajiao-hjyp · 2023-04-25T15:43:38Z

docs/guides/model_convert/convert_from_pytorch/api_difference/linalg/torch.lu.md

+| ------------- | ------------ | ------------------------------------------------------ |
+| <font color='red'> A </font>         | <font color='red'> x </font>            | 输入的 Tensor ，仅参数名不同。                                     |
+| <font color='red'> pivots </font>    | <font color='red'> pivot </font>        | 输入的 bool ，仅参数名不同。                                     |
+| <font color='red'> get_infos </font> | <font color='red'> get_infos </font>    | 输入的 bool ，仅参数名不同。                                     |


这里参数名是相同的~

huajiao-hjyp · 2023-04-25T15:44:47Z

docs/guides/model_convert/convert_from_pytorch/api_difference/linalg/torch.lu.md

+torch.lu(torch.tensor([[1.0, 2.0], [3.0, 4.0], [5.0, 6.0]]), get_infos=True, out=(A_LU, pivots, info))
+
+# Paddle 写法
+lu, p, info = paddle.linalg.lu(paddle.to_tensor([[1.0, 2.0], [3.0, 4.0], [5.0, 6.0]]).astype('float64'), get_infos=True)


可以用paddle.assign实现赋值吗

paddle.assign(x, output)中output为Tensor，此处assign会报TypeError: Unsupport paddle.assign([Variable, Variable...]) with non-scalar variable；
paddle.linalg.lu返回值为tuple,且tuple内Tensor维度不一致, 因此无法将结果转为list等方式拼接；
尚未找到合适赋值方式，暂未修改

huajiao-hjyp · 2023-04-25T15:46:48Z

docs/guides/model_convert/convert_from_pytorch/api_difference/linalg/torch.lu_unpack.md

+| <font color='red'> LU_data </font>        | <font color='red'> x </font>              | 输入的 Tensor ，仅参数名不同。                                     |
+| <font color='red'> LU_pivots </font>      | <font color='red'> y </font>              | 输入的 Tensor ，仅参数名不同。                                     |
+| <font color='red'> unpack_data </font>    | <font color='red'> unpack_ludata </font>  | 输入的 bool ，仅参数名不同。                                     |
+| <font color='red'> unpack_pivots </font>  | <font color='red'> unpack_pivots </font>  | 输入的 bool ，仅参数名不同。             |


参数名是一样的

huajiao-hjyp · 2023-04-25T15:47:43Z

docs/guides/model_convert/convert_from_pytorch/api_difference/linalg/torch.lu_unpack.md

+torch.lu_unpack(*torch.lu(torch.tensor([[1.0, 2.0], [3.0, 4.0], [5.0, 6.0]])), out=(P, L, U))
+
+# Paddle 写法
+P,L,U = paddle.linalg.lu_unpack(*paddle.linalg.lu(paddle.to_tensor([[1.0, 2.0], [3.0, 4.0], [5.0, 6.0]]).astype('float64')))


试一下可以用paddle.assign实现吗

同paddle.linalg.lu

huajiao-hjyp · 2023-04-25T15:50:01Z

docs/guides/model_convert/convert_from_pytorch/api_difference/linalg/torch.matrix_power.md

+torch.matrix_power(torch.tensor([[1., 2., 3.],[1., 4., 9.],[1., 8., 27.]]), 2, out=y)
+
+# Paddle 写法
+paddle.assign(paddle.linalg.matrix_power(paddle.to_tensor([[1, 2, 3],[1, 4, 9],[1, 8, 27]], dtype='float64'), 2), y)


这里paddle的写法中，可以将1换成 1. ，这样就不用显式写dtype了

huajiao-hjyp · 2023-04-25T15:50:46Z

docs/guides/model_convert/convert_from_pytorch/api_difference/linalg/torch.matrix_rank.md

@@ -0,0 +1,30 @@
+## [torch 参数更多 ]torch.matrix_rank


签名和api名不一致~

huajiao-hjyp · 2023-04-25T15:51:33Z

docs/guides/model_convert/convert_from_pytorch/api_difference/linalg/torch.matrix_rank.md

+| PyTorch       | PaddlePaddle | 备注                                                   |
+| ------------- | ------------ | ------------------------------------------------------ |
+| A             | x            | 输入的 Tensor ，仅参数名不同。                         |
+| tol           | tol          | 输入的 Tensor 或者 float，仅参数名不同。                         |


参数名是一致的

huajiao-hjyp · 2023-04-25T15:51:44Z

docs/guides/model_convert/convert_from_pytorch/api_difference/linalg/torch.matrix_rank.md

+| ------------- | ------------ | ------------------------------------------------------ |
+| A             | x            | 输入的 Tensor ，仅参数名不同。                         |
+| tol           | tol          | 输入的 Tensor 或者 float，仅参数名不同。                         |
+| hermitian     | hermitian    | 输入的 bool ，仅参数名不同。                            |


参数名是一致的

huajiao-hjyp · 2023-04-25T15:52:28Z

docs/guides/model_convert/convert_from_pytorch/api_difference/linalg/torch.matrix_rank.md

@@ -0,0 +1,30 @@
+## [torch 参数更多 ]torch.matrix_rank
+### [torch.matrix_rank](https://pytorch.org/docs/stable/generated/torch.linalg.matrix_rank.html?highlight=matrix_rank#torch.linalg.matrix_rank)


这里少了linalg 吧

原在线文档中该函数为torch.matrix_rank，pytorch1.13中实际只有torch.linalg.matrix_rank，已修改该api名及其文件名

huajiao-hjyp · 2023-04-25T15:53:53Z

docs/guides/model_convert/convert_from_pytorch/api_difference/ops/torch.dot.md

+torch.dot(torch.tensor([2, 3]), torch.tensor([2, 1]), out=y)
+
+# Paddle 写法
+y = paddle.dot(paddle.to_tensor([2, 3]), paddle.to_tensor([2, 1]))


赋值可以用paddle.assign

huajiao-hjyp · 2023-04-27T01:17:00Z

.../guides/model_convert/convert_from_pytorch/api_difference/linalg/torch.linalg.matrix_rank.md

+| PyTorch       | PaddlePaddle | 备注                                                   |
+| ------------- | ------------ | ------------------------------------------------------ |
+| A             | x            | 输入的 Tensor ，仅参数名不同。                         |
+| tol           | tol          | 输入的 Tensor 或者 float，仅参数名不同。                         |


参数名是一样的

huajiao-hjyp · 2023-04-27T01:17:58Z

docs/guides/model_convert/convert_from_pytorch/api_difference/linalg/torch.lu.md

@@ -0,0 +1,34 @@
+## [torch 参数更多 ]torch.lu
+
+### [torch.lu](https://pytorch.org/docs/stable/generated/torch.inverse.html?highlight=inverse#torch.inverse)


链接不正确

huajiao-hjyp · 2023-04-27T01:21:02Z

docs/guides/model_convert/convert_from_pytorch/api_difference/linalg/torch.matrix_power.md

+| PyTorch       | PaddlePaddle | 备注                                                   |
+| ------------- | ------------ | ------------------------------------------------------ |
+| input         | x            | 输入的 Tensor ，仅参数名不同。                             |
+| n             | n            | 输入的 int ，仅参数名不同。                             |


参数名相同~

YiZoey · 2023-04-27T04:14:53Z

均已更正，并重新检查类似问题，确认无误

huajiao-hjyp

LGTM

huajiao-hjyp · 2023-04-28T10:00:49Z

可以把缺失的 api 名称补充在 comment 中
torch.inverse torch.lu torch.lu_unpack torch.matrix_power 这几个文件可以放在ops下，因为我们是以torch的标准进行分类的~

YiZoey · 2023-04-29T02:42:16Z

缺失的api为torch.matmul，其已在项目仓库中存在，无需修改或添加。
文件已经迁移，感谢提醒。

zhwesky2010

辛苦改一下

zhwesky2010 · 2023-05-08T09:49:05Z

docs/guides/model_convert/convert_from_pytorch/api_difference/ops/torch.inner.md

+#### out：指定输出
+```python
+# Pytorch 写法
+torch.inner(torch.arange(1., 7.).reshape((2, 3)), torch.arange(1., 10.).reshape((3, 3)), out=y)


后面可以尽量把示例写的简单些，直接写list表示Tensor也行，这个本质是伪代码的形式让用户一眼看了就懂转换思路的。写出核心转换思路就可以，辅助运行的代码应省尽省。

zhwesky2010 · 2023-05-08T09:51:57Z

docs/guides/model_convert/convert_from_pytorch/api_difference/ops/torch.lu.md

+torch.lu(torch.tensor([[1.0, 2.0], [3.0, 4.0], [5.0, 6.0]]), get_infos=True, out=(A_LU, pivots, info))
+
+# Paddle 写法
+lu, p, info = paddle.linalg.lu(paddle.to_tensor([[1.0, 2.0], [3.0, 4.0], [5.0, 6.0]]), get_infos=True)


A_LU, pivots, info

zhwesky2010 · 2023-05-08T09:54:11Z

docs/guides/model_convert/convert_from_pytorch/api_difference/ops/torch.ger.md

+#### out：指定输出
+```python
+# Pytorch 写法
+torch.ger(torch.arange(1., 5.), torch.arange(1., 4.), out=y)


写的比较复杂，建议写简略一些

YiZoey · 2023-05-10T05:12:15Z

辛苦改一下

已修改相关样例表示

zhwesky2010

LGTM

luotao1 · 2023-05-16T07:15:01Z

hi, @YiZoey

非常感谢你对飞桨框架的贡献，我们正在运营一个PFCC组织，会通过定期分享技术知识与发布开发者主导任务的形式持续为飞桨框架做贡献，详情可见 https://github.com/luotao1 主页说明。
如果你对PFCC有兴趣，请发送邮件至 ext_paddle_oss@baidu.com，我们会邀请你加入~

映射文档 No.48

80a3be3

paddle-bot bot added contributor status: proposed labels Apr 25, 2023

YiZoey changed the title ~~映射文档 No.48~~ workl in process: 映射文档 No.48 Apr 25, 2023

task 48

5d8ec1b

YiZoey changed the title ~~workl in process: 映射文档 No.48~~ 映射文档 No.48 Apr 25, 2023

luotao1 assigned zhwesky2010 and huajiao-hjyp Apr 25, 2023

huajiao-hjyp reviewed Apr 25, 2023

View reviewed changes

YiZoey added 2 commits April 26, 2023 01:38

reg

16a0e25

reg

cac4c39

huajiao-hjyp mentioned this pull request Apr 26, 2023

PyTorch 1.13 与 Paddle 2.4 API 映射文档开源任务 #5691

Closed

huajiao-hjyp reviewed Apr 27, 2023

View reviewed changes

reg

aaf9bf9

reg

5a6e9ad

huajiao-hjyp approved these changes Apr 28, 2023

View reviewed changes

reg

3269bf7

huajiao-hjyp approved these changes May 4, 2023

View reviewed changes

zhwesky2010 reviewed May 8, 2023

View reviewed changes

reg

31e6400

reg

f4fa63a

zhwesky2010 approved these changes May 12, 2023

View reviewed changes

zhwesky2010 merged commit 4b757b3 into PaddlePaddle:develop May 12, 2023

luotao1 added the HappyOpenSource 快乐开源活动issue与PR label Aug 15, 2023

		@@ -0,0 +1,30 @@
		## [torch 参数更多 ]torch.matrix_rank
		### [torch.matrix_rank](https://pytorch.org/docs/stable/generated/torch.linalg.matrix_rank.html?highlight=matrix_rank#torch.linalg.matrix_rank)

		@@ -0,0 +1,34 @@
		## [torch 参数更多 ]torch.lu

		### [torch.lu](https://pytorch.org/docs/stable/generated/torch.inverse.html?highlight=inverse#torch.inverse)

映射文档 No.48 #5840

映射文档 No.48 #5840

Conversation

YiZoey commented Apr 25, 2023

paddle-bot bot commented Apr 25, 2023

YiZoey commented Apr 25, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

YiZoey Apr 26, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

YiZoey Apr 26, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

YiZoey commented Apr 27, 2023

huajiao-hjyp left a comment

Choose a reason for hiding this comment

huajiao-hjyp commented Apr 28, 2023

YiZoey commented Apr 29, 2023

zhwesky2010 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

YiZoey commented May 10, 2023

zhwesky2010 left a comment

Choose a reason for hiding this comment

luotao1 commented May 16, 2023

YiZoey commented Apr 25, 2023 •

edited

Loading

YiZoey Apr 26, 2023 •

edited

Loading

YiZoey Apr 26, 2023 •

edited

Loading