【PaddlePaddle Hackathon 4 NO.23】为 Paddle 新增 vander API #51048

Li-fAngyU · 2023-03-01T04:26:48Z

PR types

New features

PR changes

APIs

Describe

rfc文档链接：PaddlePaddle/community#386
rfc 修复链接：PaddlePaddle/community#464
中文文档：PaddlePaddle/docs#5681

为Paddle新增Vander API

paddle-bot · 2023-03-01T04:26:51Z

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.

luotao1

请补充下对应的中文文档

luotao1 · 2023-03-06T03:08:53Z

python/paddle/tensor/math.py

+    is named after Alexandre-Theophile Vandermonde.
+
+    Args:
+        x (Tensor): The input tensor, it must be 1-D Tensor, and it's data type should be ['complex64', 'complex128', 'float32', 'float64', 'int32', 'int64'].


'complex64', 'complex128', 如果支持复数的话，单测以及示例代码中均要体现

luotao1 · 2023-03-06T03:12:13Z

python/paddle/fluid/tests/unittests/test_vander.py

+                    res[0].size, out_ref.size, rtol=1e-05
+                )
+            else:
+                assert res[0] is None


52-57行，为什么cpu/gpu模式下不一样呢？

这是因为在静态图模式下，paddle.empty 的返回值有点区别，我不太清楚是不是bug就做了点处理。

example code:

import paddle print(paddle.__version__) import numpy as np N = 0 places = [paddle.CPUPlace(),paddle.CUDAPlace(0)] paddle.enable_static() for place in places: with paddle.static.program_guard(paddle.static.Program()): out = paddle.empty([3, N]) exe = paddle.static.Executor(place) res = exe.run(fetch_list=[out]) print('static:', place, res) # 2.3.2 # static: Place(cpu) [None] # static: Place(gpu:0) [array([], shape=(3, 0), dtype=float32)]

Li-fAngyU · 2023-03-06T08:54:45Z

已补充复数的单测代码，以及示例代码。

注：因静态图下, paddle.empty 不支持创建数据类型为complex 的Tensor，因此单测中没有包括静态图下测试 paddle.vander 是否支持复数类型的代码。

example code:

import paddle
print(paddle.__version__)
places = [paddle.CPUPlace()]
paddle.enable_static()
for place in places:
    with paddle.static.program_guard(paddle.static.Program()):
        out = paddle.empty([3, 0],dtype=paddle.complex64)
        exe = paddle.static.Executor(place)
        res = exe.run(fetch_list=[out])
# 0.0.0
#       5 for place in places:
#       6     with paddle.static.program_guard(paddle.static.Program()):
# ----> 7         out = paddle.empty([3, 0],dtype=paddle.complex64)
#       8         exe = paddle.static.Executor(place)
#       9         res = exe.run(fetch_list=[out])
# TypeError: The data type of 'dtype' in empty must be ['bool', 'float16', 'float32', 'float64', 'int32', 'int64'], but received complex64.

luotao1 · 2023-03-07T11:08:11Z

因静态图下，paddle.empty 不支持创建数据类型为complex 的Tensor

Paddle/python/paddle/tensor/creation.py

Lines 1788 to 1792 in 771b589

    
           check_dtype( 
        
               dtype, 
        
               'dtype', 
        
               ['bool', 'float16', 'float32', 'float64', 'int32', 'int64'], 
        
               'empty',

1791行加上complex64和complex128就可以了，可以单独提一个PR修一下。

luotao1 · 2023-03-07T11:28:23Z

python/paddle/tensor/math.py

+    if N < 0:
+        raise ValueError("N must be non-negative.")
+
+    res = paddle.empty([x.shape[0], N], dtype=x.dtype)


这是因为在静态图模式下，paddle.empty 的返回值有点区别，我不太清楚是不是bug就做了点处理。#51048 (comment)

我能在develop分支复现，也看到你的issue了，但什么情况下N会等于0呢？

N (int): Number of columns in the output. If N is not specified, a square array is returned (N = len(x)).

请问Paddle静态图下因设备不同导致 empty API在返回空Tensor时不一致是正常情况吗？ #51133

因torch，和numpy的vander都能支持N=0这一极端条件（torch的话是支持空tensor，以前的paddle版本是不支持空tensor，然后我就在单测中多考虑这一情况了）理论上N应该是大于0的

>>> import torch >>> x = torch.tensor([1, 2, 3, 5]) >>> torch.vander(x, N=0) tensor([], size=(4, 0), dtype=torch.int64) >>> import paddle >>> paddle.to_tensor([4,0]) Tensor(shape=[2], dtype=int64, place=Place(cpu), stop_gradient=True, [4, 0])

请 @cuicheng01 帮忙看下【PaddlePaddle Hackathon 4 NO.23】为 Paddle 新增 vander API #51048 (comment) 如何解决

@Li-fAngyU N=0的时候，能否用paddle.to_tensor呢？

好的。

不是很理解 "N=0的时候，能否用paddle.to_tensor?"具体是什么意思, 是指能否用paddle.to_tensor创建空Tensor吗？
（如果用paddle.to_tensor创建空Tensor a的话，a的shape 好像就没法构造成(4,0)的形式了）

不好意思，paddle.to_tensor确实无法构造，看下paddle.rand系列

>>> import paddle >>> paddle.randint(0, 1, [4, 0]) Tensor(shape=[4, 0], dtype=int64, place=Place(cpu), stop_gradient=True, [[], [], [], []])

抱歉，paddle.to_tensor 是可以构建空Tensor的，shape的话直接用reshape就可以了。

>>> a = paddle.to_tensor([]).reshape([3,0]) >>> a Tensor(shape=[3, 0], dtype=float32, place=Place(cpu), stop_gradient=True, [[], [], []])

所以目前要先在N=0的时候，用paddle.to_tensor去替代paddle.empty吗？

@Li-fAngyU paddle.to_tensor在静态图下不能构造含有0 dim的tensor，建议还是使用paddle.empty

import paddle print(paddle.__version__) import numpy as np N = 0 #places = [paddle.CPUPlace(),paddle.CUDAPlace(0)] places = [paddle.CPUPlace()] paddle.enable_static() for place in places: with paddle.static.program_guard(paddle.static.Program()): #out = paddle.empty([3, N]) out = paddle.to_tensor([]).reshape([3,N]) exe = paddle.static.Executor(place) res = exe.run(fetch_list=[out]) print('static:', place, res) ###### λ ffd50c7f717b /Paddle {develop} python3.7 test.py grep: warning: GREP_OPTIONS is deprecated; please use an alias or script 0.0.0 Traceback (most recent call last): File "test.py", line 11, in <module> out = paddle.to_tensor([]).reshape([3,N]) File "/usr/local/python3.7.0/lib/python3.7/site-packages/paddle/tensor/manipulation.py", line 3588, in reshape attrs["shape"] = get_attr_shape(shape) File "/usr/local/python3.7.0/lib/python3.7/site-packages/paddle/tensor/manipulation.py", line 3571, in get_attr_shape % (dim_idx, len(x.shape)) AssertionError: The index of 0 in `shape` must be less than the input tensor X's dimensions. But received shape[1] = 0, X's dimensions = 1.

和 @cuicheng01 @zhouwei25 讨论，应该是静态图CPUPlace的问题，换成 paddle.randint(0, 1, [4, 0])，输出也是None

static: Place(cpu) [None]

Li-fAngyU · 2023-03-09T09:56:27Z

@cuicheng01

Li-fAngyU · 2023-03-20T03:59:07Z

@luotao1 请问PR-CR-Build里报错 make[2]: *** No rule to make target '../python/paddle/fluid/tests/unittests/test_vander.py', needed by 'python/build/.timestamp'. Stop. 该怎么解决呢？

luotao1 · 2023-03-20T09:07:42Z

该怎么解决呢

已集中整理到 #51195 (comment)

luotao1 · 2023-03-21T09:51:52Z

@Li-fAngyU paddle-build失败的问题已修复，可以rerun一下

fix build_ci_problem #51910

Li-fAngyU · 2023-03-22T01:05:42Z

收到

luotao1

LGTM

jeff41404

LGTM

sunzhongkai588 · 2023-03-29T03:19:10Z

python/paddle/tensor/math.py

+
+            x = paddle.to_tensor([1., 2., 3.], dtype="float32")
+            out = paddle.vander(x)
+            print(out.numpy())


print输出一定要在后面加 .numpy()吗？去除后能否正常输出

加了.numpy() 是为了让输出更简洁一点，去除后可以正常输出。

sunzhongkai588 · 2023-03-29T03:21:02Z

python/paddle/tensor/math.py

+        increasing(bool): Order of the powers of the columns. If True, the powers increase from left to right, if False (the default) they are reversed.
+        name (str, optional): For details, please refer to :ref:`api_guide_Name`. Generally, no setting is required. Default: None.
+    Returns:
+        Vandermonde matrix with shape: (len(x), N). If increasing is False, the first column is x^(N-1), the second x^(N-2) and so forth. If increasing is True, the columns are x^0, x^1, ..., x^(N-1).


x^(N-1) 这类带次方的数字写成公式叭, 即写成 :math:xxxxx

返回值请按规范写，需要先描述 API 返回值的类型，然后再描述 API 的返回值及其含义。

shape后不加冒号。returns的描述里避免出现冒号，因为冒号前面的会解析为 Returns type

paddle-bot · 2023-03-31T01:16:20Z

很抱歉，经过我们的反复讨论，你的PR暂未达到合入标准，请阅读飞桨原生算子开发规范，你可以重新提交新的PR，我们先将此PR关闭，感谢你的贡献。
Sorry to inform you that through our discussion, your PR fails to meet the merging standard (Reference: Paddle Custom Operator Design Doc). You can also submit an new one. Thank you.

… api_vander

sunzhongkai588

LGTM

first commit of vander api

6bf60e0

paddle-bot bot added contributor External developers status: proposed labels Mar 1, 2023

luotao1 assigned luotao1 and cloud2009 Mar 1, 2023

fix file name

040457f

Li-fAngyU mentioned this pull request Mar 1, 2023

【PaddlePaddle Hackathon 第四期】任务总览 #50629

Closed

Li-fAngyU added 7 commits March 1, 2023 18:59

fix code style problem

fc85609

fix error of vander api at math.py

200ae0d

fix code style problem

ce5a955

fix code style

e2d1f97

code style

b4bea35

code style

68942cb

fix test vander api error!

368adf2

luotao1 reviewed Mar 6, 2023

View reviewed changes

Li-fAngyU and others added 3 commits March 6, 2023 15:34

Merge branch 'PaddlePaddle:develop' into api_vander

29b84b0

add complex test!

eba525a

add complex example code of vander.

9f4ff05

luotao1 assigned Ligoml Mar 6, 2023

Li-fAngyU mentioned this pull request Mar 6, 2023

【PaddlePaddle Hackathon 4 No.23】新增 API vander PaddlePaddle/docs#5681

Merged

fix sample code error

9c8659b

Ligoml mentioned this pull request Mar 7, 2023

【PaddlePaddle Hackathon 第四期】任务总览 #51281

Closed

luotao1 reviewed Mar 7, 2023

View reviewed changes

Li-fAngyU mentioned this pull request Mar 8, 2023

update empty api to support complex dtype at static mode #51377

Merged

luotao1 added the API label Mar 10, 2023

paddle-bot bot removed the status: proposed label Mar 10, 2023

Li-fAngyU reopened this Mar 17, 2023

fix error in test_vander

407893e

luotao1 mentioned this pull request Mar 20, 2023

【Hackathon 4th 】add Trapezoid API && add Cumulative_trapezoid API #51195

Merged

Li-fAngyU and others added 3 commits March 22, 2023 09:06

Merge branch 'PaddlePaddle:develop' into api_vander

acf59bd

change paddle.fluid to paddle.static in test_vander file.

1f29e36

fix params

7196e17

luotao1 previously approved these changes Mar 24, 2023

View reviewed changes

jeff41404 previously approved these changes Mar 27, 2023

View reviewed changes

luotao1 requested a review from sunzhongkai588 March 28, 2023 02:57

luotao1 assigned sunzhongkai588 Mar 28, 2023

sunzhongkai588 reviewed Mar 29, 2023

View reviewed changes

Li-fAngyU closed this Mar 31, 2023

Li-fAngyU force-pushed the api_vander branch from 7196e17 to 9fd4fd5 Compare March 31, 2023 01:16

paddle-bot bot added the status: not progressed label Mar 31, 2023

Li-fAngyU added 3 commits March 31, 2023 09:24

Merge branch 'api_vander' of https://github.com/Li-fAngyU/Paddle into…

03bb29a

… api_vander

update vander return describe

3507d28

vander examples.

956a36b

Li-fAngyU reopened this Mar 31, 2023

Li-fAngyU dismissed stale reviews from jeff41404 and luotao1 via 956a36b March 31, 2023 01:30

code style

5d748f1

sunzhongkai588 approved these changes Mar 31, 2023

View reviewed changes

luotao1 merged commit 7b53923 into PaddlePaddle:develop Mar 31, 2023

Li-fAngyU deleted the api_vander branch July 6, 2023 02:17

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

【PaddlePaddle Hackathon 4 NO.23】为 Paddle 新增 vander API #51048

【PaddlePaddle Hackathon 4 NO.23】为 Paddle 新增 vander API #51048

Li-fAngyU commented Mar 1, 2023 •

edited by luotao1

paddle-bot bot commented Mar 1, 2023

luotao1 left a comment

luotao1 Mar 6, 2023

Li-fAngyU Mar 6, 2023

luotao1 Mar 6, 2023

Li-fAngyU Mar 6, 2023 •

edited

Li-fAngyU commented Mar 6, 2023 •

edited

luotao1 commented Mar 7, 2023 •

edited

luotao1 Mar 7, 2023 •

edited

Li-fAngyU Mar 8, 2023

luotao1 Mar 9, 2023

Li-fAngyU Mar 9, 2023 •

edited

luotao1 Mar 9, 2023

Li-fAngyU Mar 9, 2023 •

edited

cuicheng01 Mar 13, 2023 •

edited by luotao1

luotao1 Mar 13, 2023

Li-fAngyU commented Mar 9, 2023

Li-fAngyU commented Mar 20, 2023

luotao1 commented Mar 20, 2023

luotao1 commented Mar 21, 2023

Li-fAngyU commented Mar 22, 2023

luotao1 left a comment

jeff41404 left a comment

sunzhongkai588 Mar 29, 2023

Li-fAngyU Mar 31, 2023

sunzhongkai588 Mar 29, 2023

Li-fAngyU Mar 31, 2023

paddle-bot bot commented Mar 31, 2023

sunzhongkai588 left a comment

【PaddlePaddle Hackathon 4 NO.23】为 Paddle 新增 vander API #51048

【PaddlePaddle Hackathon 4 NO.23】为 Paddle 新增 vander API #51048

Conversation

Li-fAngyU commented Mar 1, 2023 • edited by luotao1

PR types

PR changes

Describe

paddle-bot bot commented Mar 1, 2023

luotao1 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

Li-fAngyU Mar 6, 2023 • edited

Choose a reason for hiding this comment

Li-fAngyU commented Mar 6, 2023 • edited

luotao1 commented Mar 7, 2023 • edited

luotao1 Mar 7, 2023 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Li-fAngyU Mar 9, 2023 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Li-fAngyU Mar 9, 2023 • edited

Choose a reason for hiding this comment

cuicheng01 Mar 13, 2023 • edited by luotao1

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Li-fAngyU commented Mar 9, 2023

Li-fAngyU commented Mar 20, 2023

luotao1 commented Mar 20, 2023

luotao1 commented Mar 21, 2023

Li-fAngyU commented Mar 22, 2023

luotao1 left a comment

Choose a reason for hiding this comment

jeff41404 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

Choose a reason for hiding this comment

paddle-bot bot commented Mar 31, 2023

sunzhongkai588 left a comment

Choose a reason for hiding this comment

Li-fAngyU commented Mar 1, 2023 •

edited by luotao1

Li-fAngyU Mar 6, 2023 •

edited

Li-fAngyU commented Mar 6, 2023 •

edited

luotao1 commented Mar 7, 2023 •

edited

luotao1 Mar 7, 2023 •

edited

Li-fAngyU Mar 9, 2023 •

edited

Li-fAngyU Mar 9, 2023 •

edited

cuicheng01 Mar 13, 2023 •

edited by luotao1