【Hackathon 5th No.17】为 Paddle 新增 pdist API -part by cocoshe · Pull Request #57869 · PaddlePaddle/Paddle

cocoshe · 2023-10-03T14:28:47Z

PR types

New features

PR changes

APIs

Description

RFC：

中文文档：

【Hackathon 5th No.17】为 Paddle 新增 pdist API 中文文档 docs#6391

paddle-bot · 2023-10-03T14:28:52Z

你的PR提交成功，感谢你对开源项目的贡献!
请关注后续CI自动化测试结果，详情请参考Paddle-CI手册。
Your PR has been submitted. Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

paddle-ci-bot · 2023-10-15T03:09:46Z

Sorry to inform you that 9c651d9's CIs have passed for more than 7 days. To prevent PR conflicts, you need to re-run all CIs manually.

paddle-ci-bot · 2023-10-29T03:12:22Z

Sorry to inform you that 13d04ad's CIs have passed for more than 7 days. To prevent PR conflicts, you need to re-run all CIs manually.

cocoshe · 2023-11-16T14:52:40Z

@zxcd 可以辛苦帮忙review一下嘛？看看有没有啥问题~

zxcd · 2023-11-17T07:18:51Z

+        paddle.enable_static()
+
+
+class TestpdistAPICase1(TestpdistAPI):


规范一下测试命名

zxcd · 2023-11-17T07:19:20Z

+            exe = paddle.static.Executor(self.place)
+            res = exe.run(feed={'x': self.x}, fetch_list=[out])
+            out_ref = ref_pdist(self.x, self.p)
+            np.testing.assert_allclose(out_ref, res[0], rtol=1e-5, atol=1e-5)


这里应该不需要使用rtol, atol?

>>> x = np.random.rand(3, 4).astype('float64') >>> t_x = paddle.to_tensor(x) >>> x == t_x.numpy() array([[ True, True, True, True], [ True, True, True, True], [ True, True, True, True]]) >>> np_norm = np.linalg.norm(x, axis=-1) >>> pd_norm = paddle.linalg.norm(t_x, axis=-1) >>> np_norm == pd_norm.numpy() array([ True, True, True]) >>> x = np.random.rand(5, 6).astype('float64') >>> t_x = paddle.to_tensor(x) >>> np_norm = np.linalg.norm(x, axis=-1) >>> pd_norm = paddle.linalg.norm(t_x, axis=-1) >>> np_norm == pd_norm.numpy() array([ True, True, True, True, True]) >>> x = np.random.rand(5, 6).astype('float64') >>> t_x = paddle.to_tensor(x) >>> np_norm = np.linalg.norm(x, axis=-1) >>> pd_norm = paddle.linalg.norm(t_x, axis=-1) >>> np_norm == pd_norm.numpy() array([ True, True, True, True, False]) >>> >>> x = np.random.rand(5, 6).astype('float64') >>> t_x = paddle.to_tensor(x) >>> np_norm = np.linalg.norm(x, axis=-1) >>> pd_norm = paddle.linalg.norm(t_x, axis=-1) >>> np_norm == pd_norm.numpy() array([ True, True, True, True, False]) >>> >>> x = np.random.rand(5, 6).astype('float64') >>> t_x = paddle.to_tensor(x) >>> np_norm = np.linalg.norm(x, axis=-1) >>> pd_norm = paddle.linalg.norm(t_x, axis=-1) >>> np_norm == pd_norm.numpy() array([ True, True, True, True, False]) >>> >>> x = np.random.rand(5, 6).astype('float64') >>> t_x = paddle.to_tensor(x) >>> np_norm = np.linalg.norm(x, axis=-1) >>> pd_norm = paddle.linalg.norm(t_x, axis=-1) >>> np_norm == pd_norm.numpy() array([False, False, False, True, True]) >>> >>> x = np.random.rand(5, 6).astype('float64') >>> t_x = paddle.to_tensor(x) >>> np_norm = np.linalg.norm(x, axis=-1) >>> pd_norm = paddle.linalg.norm(t_x, axis=-1) >>> np_norm == pd_norm.numpy() array([ True, True, False, False, True]) >>>

如果不用rtol atol的话精度过不了。我进行了一些尝试，发现norm看样子像是没和numpy对齐，在cdist的单测中也是放宽了精度

Paddle/test/legacy_test/test_cdist.py

Line 56 in 907e425

np.testing.assert_allclose(out_ref, res[0], rtol=1e-5, atol=1e-5)

zxcd · 2023-11-17T07:49:24Z

+        compute_mode (str, optional): The mode for compute distance.
+
+            - ``use_mm_for_euclid_dist_if_necessary`` , for p = 2.0 and (P > 25 or R > 25), it will use matrix multiplication to calculate euclid distance if possible.
+            - ``use_mm_for_euclid_dist`` , for p = 2.0, it will use matrix multiplication to calculate euclid distance.


仅针对pdist这个API来说，cdist中使用到的这几个mode是否会用到？如果完全用不到我理解这个地方可能不需要添加mode参数。

仅针对pdist这个API来说，cdist中使用到的这几个mode是否会用到？如果完全用不到我理解这个地方可能不需要添加mode参数。

我这里是把pdist当作一种特殊的cdist来实现：cdist输入x:[P,M], y:[R,M], 输出shape: [P,R]，所以我在这里就是通过计算两份相同输入的cdist来实现pdist（令y=x），也就是cdist(x, x), 输入shape:[N, M]，输出shape:[N, N]，然后取上三角（去除对角线）得到结果。

对照起来看的话，cdist中的P和R对应现在输入[N,M]中的N，那当N较大的时候（大于25）我感觉是不是也可以采取cdist中的矩阵策略，当然不采用矩阵方法的话，应该就是直接norm然后取三角了，不知道这么想对不对

从竞品对照来看，pdist的结果都是比较统一的，不存在N大于25时需要进行额外操作。另外从公式来说cdist和pdist的意义并不完全一致，我建议与竞品保持一致，不需要额外添加该参数。

从竞品对照来看，pdist的结果都是比较统一的，不存在N大于25时需要进行额外操作。我建议与竞品保持一致，不需要额外添加该参数。

OKK，辛苦review，我之后稍作修改~

paddle-ci-bot · 2023-11-19T03:21:40Z

Sorry to inform you that 8bcbe47's CIs have passed for more than 7 days. To prevent PR conflicts, you need to re-run all CIs manually.

… pdist_coco_dev

zxcd · 2023-11-30T08:24:56Z

+        self.x = np.random.rand(50, 20).astype('float64')
+
+
+class TestPdistAPICase8(TestPdistAPI):


建议给出有意义的测试命名

zxcd · 2023-11-30T08:37:32Z

+        .. code-block:: python
+
+            >>> import paddle
+            >>> a = paddle.randn([4, 5])


doc中给出seed，不然PR-CI-Static-Check过不了
参考：

Paddle/python/paddle/tensor/linalg.py

Line 764 in 855e51e

>>> paddle.seed(2023)

doc中给出seed，不然PR-CI-Static-Check过不了参考：

Paddle/python/paddle/tensor/linalg.py

Line 764 in 855e51e

>>> paddle.seed(2023)

现在添加了，但是好像没作用嘛？

添加了seed之后，你的print的结果也会有变化，这块的输出你可以参考报错的内容

添加了seed之后，你的print的结果也会有变化，这块的输出你可以参考报错的内容

明白了，稍后修改~

… pdist_coco_dev

…to pdist_coco_dev

luotao1 · 2023-12-06T06:29:53Z

2023-12-06 09:50:20 ----------------Check results--------------------
2023-12-06 09:50:20 >>> Sample code test capacity: {'\''cpu'\''}
2023-12-06 09:50:20 >>> 1 sample codes ran failed in env: {'\''cpu'\''}
2023-12-06 09:50:20 <DocTest(<modname?> paddle.nn.functional.pdist:1:0 ln 1)>, running time: 0.974s
2023-12-06 09:50:20 >>> Mistakes found in sample codes in env: {'\''cpu'\''}!

zxcd

LGTM

jeff41404 · 2023-12-12T04:48:07Z

    'conv3d',
    'conv3d_transpose',
    'pairwise_distance',
+    'pdist',


which path do we recommend users to use? If paddle.pdist is recommended, it cannot be added to this __all__ list. if paddle.nn.functional.pdist, it cannot be added to the __all__ list in python/paddle/__init__.py

Tks for you review, I think I prefer paddle.pdist path, I removed this line then.

jeff41404

LGTM

luotao1 · 2023-12-13T06:34:12Z

对应中文文档可以提上来

Co-authored-by: zachary sun <70642955+sunzhongkai588@users.noreply.github.com>

sunzhongkai588

LGTM

paddle-bot Bot added the contributor External developers label Oct 3, 2023

add pdist api

3e19c90

Ligoml mentioned this pull request Oct 7, 2023

【PaddlePaddle Hackathon 5th】开源贡献个人挑战赛 #57262

Closed

cocoshe added 3 commits October 8, 2023 00:53

move pdist to nn.functional, expose paddle.pdist api

51b908a

clean

ced239c

clean

9c651d9

luotao1 added the PaddlePaddle Hackathon label Oct 9, 2023

luotao1 assigned luotao1 and zxcd Oct 9, 2023

fix codestyle

13d04ad

fix conflict

8bcbe47

zxcd reviewed Nov 17, 2023

View reviewed changes

cocoshe added 3 commits November 24, 2023 12:54

Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into…

212bdf7

… pdist_coco_dev

remove compute_mode

7bbc22e

for api name rules

20f82de

cocoshe requested a review from zxcd November 24, 2023 18:40

Merge branch 'develop' into pdist_coco_dev

b06dd06

zxcd reviewed Nov 30, 2023

View reviewed changes

cocoshe and others added 6 commits November 30, 2023 16:43

Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into…

2e259a6

… pdist_coco_dev

Update test_pdist.py

e9dfa5a

Merge branch 'pdist_coco_dev' of https://github.com/cocoshe/Paddle in…

cddec3b

…to pdist_coco_dev

add seed

23d5f7e

fix code sample

3a9fbfb

Merge branch 'pdist_coco_dev' of https://github.com/cocoshe/Paddle in…

e41f9dc

…to pdist_coco_dev

cocoshe and others added 2 commits December 5, 2023 17:17

fix doc

008ae5b

Update distance.py

114453a

cocoshe added 2 commits December 9, 2023 10:22

gpu0 to cpu in api doc

171339d

gpu0 to cpu in api doc

8a281b4

zxcd approved these changes Dec 11, 2023

View reviewed changes

luotao1 assigned jeff41404 and sunzhongkai588 Dec 11, 2023

jeff41404 reviewed Dec 12, 2023

View reviewed changes

cocoshe added 3 commits December 12, 2023 12:59

remove pdist in nn.functional __all__ list

a9d053c

Merge branch 'develop' into pdist_coco_dev

b3816df

Update __init__.py

3f7b607

jeff41404 previously approved these changes Dec 13, 2023

View reviewed changes

fix en doc

e3e5abf

cocoshe dismissed jeff41404’s stale review via e3e5abf December 13, 2023 06:35

cocoshe mentioned this pull request Dec 13, 2023

【Hackathon 5th No.17】为 Paddle 新增 pdist API 中文文档 PaddlePaddle/docs#6391

Merged

sunzhongkai588 reviewed Dec 13, 2023

View reviewed changes

Comment thread python/paddle/nn/functional/distance.py Resolved

Comment thread python/paddle/nn/functional/distance.py Resolved

cocoshe and others added 2 commits December 13, 2023 15:59

Update python/paddle/nn/functional/distance.py

8172f28

Co-authored-by: zachary sun <70642955+sunzhongkai588@users.noreply.github.com>

Update python/paddle/nn/functional/distance.py

0146a37

Co-authored-by: zachary sun <70642955+sunzhongkai588@users.noreply.github.com>

sunzhongkai588 approved these changes Dec 14, 2023

View reviewed changes

luotao1 merged commit fa440e9 into PaddlePaddle:develop Dec 14, 2023

luotao1 changed the title ~~【Hackathon 5th No.17】为 Paddle 新增 pdist API~~ 【Hackathon 5th No.17】为 Paddle 新增 pdist API -part Dec 14, 2023

cocoshe mentioned this pull request Dec 22, 2023

[WeeklyReports] 2023.12.10~2023.12.24 周报收集 PFCCLab/Starter#69

Closed

21 of 26 tasks

cocoshe mentioned this pull request Jan 19, 2024

[WeeklyReports] 2023.12.25~2024.1.19 周报收集 PFCCLab/Starter#97

Closed

18 of 24 tasks

		paddle.enable_static()


		class TestpdistAPICase1(TestpdistAPI):

		self.x = np.random.rand(50, 20).astype('float64')


		class TestPdistAPICase8(TestPdistAPI):

Conversation

cocoshe commented Oct 3, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

PR types

PR changes

Description

Uh oh!

paddle-bot Bot commented Oct 3, 2023

Uh oh!

paddle-ci-bot Bot commented Oct 15, 2023

Uh oh!

paddle-ci-bot Bot commented Oct 29, 2023

Uh oh!

cocoshe commented Nov 16, 2023

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

cocoshe Nov 17, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

zxcd Nov 22, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

paddle-ci-bot Bot commented Nov 19, 2023

Uh oh!

Choose a reason for hiding this comment

Uh oh!

zxcd Nov 30, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

luotao1 commented Dec 6, 2023

Uh oh!

zxcd left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jeff41404 left a comment

Choose a reason for hiding this comment

Uh oh!

luotao1 commented Dec 13, 2023

Uh oh!

Uh oh!

Uh oh!

sunzhongkai588 left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

cocoshe commented Oct 3, 2023 •

edited

Loading

cocoshe Nov 17, 2023 •

edited

Loading

zxcd Nov 22, 2023 •

edited

Loading

zxcd Nov 30, 2023 •

edited

Loading