Add digamma_op and unittest by zyfncg · Pull Request #33278 · PaddlePaddle/Paddle

zyfncg · 2021-06-02T03:06:59Z

PR types

New features

PR changes

OPs

Describe

Add digamma_op and unittest

API:

paddle.digamma() / paddle.tensor.digamma()

Code position:

python/paddle/tensor/math.py

Example:

import paddle

data = paddle.to_tensor([[1, 1.5], [0, -2.2]], dtype='float32')
res = paddle.digamma(data)
print(res)
# Tensor(shape=[2, 2], dtype=float32, place=CUDAPlace(0), stop_gradient=True,
#       [[-0.57721591,  0.03648996],
#        [ nan       ,  5.32286835]])

Doc:

Cn Doc PR: PaddlePaddle/docs#3574

paddle-bot-old · 2021-06-02T03:07:03Z

Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

CLAassistant · 2021-06-02T03:07:04Z

All committers have signed the CLA.

chenwhql · 2021-06-04T02:47:41Z

paddle/fluid/operators/digamma_op.cc

@@ -0,0 +1,100 @@
+/* Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.


2020 -> 2021

chenwhql · 2021-06-04T02:48:54Z

paddle/fluid/operators/digamma_op.cu

@@ -0,0 +1,64 @@
+/* Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.


chenwhql · 2021-06-04T03:02:57Z

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

+    def init_dtype_type(self):
+        self.dtype = np.float32
+
+    def test_check_grad_normal(self):


use default numeric_grad_delta is ok

chenwhql · 2021-06-04T03:06:19Z

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

+        self.check_output()
+
+    def test_check_grad_normal(self):
+        self.check_grad(['X'], 'Out', numeric_grad_delta=1e-7)


use default value

chenwhql · 2021-06-04T03:09:16Z

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

+                    out_value = exe.run(feed=input_dict, fetch_list=[out.name])
+                    self.assertEqual(
+                        np.allclose(
+                            out_value[0], sc_res, rtol=1e-04), True)


can use more samll rtol?

Done, update the rtol to 1e-05

chenwhql · 2021-06-04T03:09:22Z

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

+                with fluid.dygraph.guard(place):
+                    input_t = paddle.to_tensor(input)
+                    res = paddle.digamma(input_t).numpy()
+                    self.assertEqual(np.allclose(res, sc_res, rtol=1e-04), True)


Done, update the rtol to 1e-05

chenwhql · 2021-06-04T03:12:15Z

python/paddle/tensor/math.py

+        name(str, optional): The default value is None.  Normally there is no need for 
+            user to set this property.  For more information, please refer to :ref:`api_guide_Name`
+    Returns:
+        Tensor, the digamma of the input Tensor computed element-wise.


element-wise?

Done, has modified this sentence.

chenwhql · 2021-06-04T03:13:44Z

paddle/fluid/operators/digamma_op.cu

+
+REGISTER_OP_CUDA_KERNEL(
+    digamma_grad,
+    ops::DigammaGradKernel<paddle::platform::CUDADeviceContext, float>,


need special DigammaGradKernel here?

register a CudaKernel for digamma grad is necessary

chenwhql · 2021-06-04T03:14:59Z

paddle/fluid/operators/digamma_op.cu

+};
+
+template <typename T>
+class DigammaKernel<platform::CUDADeviceContext, T>


here is faster? maybe can test it and add some comments

Done, it isn't faster in the test, so removed it.

… zyf-addop

chenwhql · 2021-06-04T12:18:09Z

paddle/fluid/operators/digamma_op.cu

+limitations under the License. */
+
+#include <unsupported/Eigen/SpecialFunctions>
+#include "paddle/fluid/operators/digamma_op.h"


here we only need the header digamma_op.h, remove other headers

… zyf-addop

chenwhql

LGTM

…test

… zyf-addop

chenwhql

LGTM

… zyf-addop

XiaoguangHu01

LGTM

TCChenlong

LGTM

Add digamma_op and unittest

125f7fe

zyfncg and others added 2 commits June 2, 2021 12:49

add digamma_op api

78e1fd9

Merge branch 'PaddlePaddle:develop' into zyf-addop

2d6c3b9

chenwhql reviewed Jun 4, 2021

View reviewed changes

zyfncg added 3 commits June 4, 2021 06:52

remove special DigammaCudaKernel and correct some docs

31759be

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

6c45985

… zyf-addop

Merge branch 'zyf-addop' of github.com:zyfncg/Paddle into zyf-addop

47a0592

chenwhql reviewed Jun 4, 2021

View reviewed changes

zyfncg added 4 commits June 4, 2021 12:34

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

0ec2ca4

… zyf-addop

remove unused headers

37f8f81

fix api doc error

66f9ab0

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

bfc12ba

… zyf-addop

zyfncg mentioned this pull request Jun 8, 2021

Add digamma api cn doc PaddlePaddle/docs#3574

Merged

chenwhql previously approved these changes Jun 8, 2021

View reviewed changes

zyfncg added 2 commits June 10, 2021 06:48

merge the 'PaddlePaddle:develop' branch and use larger shape in unit …

58a486b

…test

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

2c3899a

… zyf-addop

zyfncg dismissed chenwhql’s stale review via 2c3899a June 10, 2021 06:49

chenwhql previously approved these changes Jun 10, 2021

View reviewed changes

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

469bd9d

… zyf-addop

zyfncg dismissed chenwhql’s stale review via 469bd9d June 11, 2021 07:02

XiaoguangHu01 approved these changes Jun 11, 2021

View reviewed changes

TCChenlong approved these changes Jun 15, 2021

View reviewed changes

chenwhql approved these changes Jun 15, 2021

View reviewed changes

chenwhql merged commit 02a6d49 into PaddlePaddle:develop Jun 15, 2021

zyfncg deleted the zyf-addop branch July 28, 2021 04:13

		@@ -0,0 +1,100 @@
		/* Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.

		@@ -0,0 +1,64 @@
		/* Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.

Conversation

zyfncg commented Jun 2, 2021 • edited by chenwhql Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

PR types

PR changes

Describe

Uh oh!

paddle-bot-old bot commented Jun 2, 2021

Uh oh!

CLAassistant commented Jun 2, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

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!

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!

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!

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!

chenwhql Jun 4, 2021 • 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!

zyfncg Jun 4, 2021 • 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!

chenwhql left a comment

Choose a reason for hiding this comment

Uh oh!

chenwhql left a comment

Choose a reason for hiding this comment

Uh oh!

XiaoguangHu01 left a comment

Choose a reason for hiding this comment

Uh oh!

TCChenlong left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

zyfncg commented Jun 2, 2021 •

edited by chenwhql

Loading

CLAassistant commented Jun 2, 2021 •

edited

Loading

chenwhql Jun 4, 2021 •

edited

Loading

zyfncg Jun 4, 2021 •

edited

Loading