【Hackathon 7th No.39】为 Paddle 代码转换工具新增 API 转换规则（第 6 组） #477

Asthestarsfalll · 2024-09-18T07:26:46Z

PR Docs

PR APIs

torch.nn.functional.lp_pool1d
torch.nn.functional.lp_pool2d
torch.nn.functional.threshold_
torch.nn.functional.feature_alpha_dropout
torch.nn.functional.scaled_dot_product_attention
torch.nn.LPPool1d
torch.nn.LPPool2d
torch.nn.Softmin
torch.nn.AdaptiveLogSoftmaxWithLoss
torch.nn.parameter.UninitializedParameter
torch.nn.parameter.UninitializedBuffer
torch.nn.CircularPad3d
torch.nn.utils.parametrizations.weight_norm
torch.optim.RAdam
torch.optim.NAdam

目前问题：

lp_pool torch行为与Paddle不一致，当norm_type为inf时，应该与max_pool一致，而torch返回1。
未找到UninitializedParameter与UninitializedBuffer的映射

paddle-bot · 2024-09-18T07:26:51Z

Thanks for your contribution!

zhwesky2010 · 2024-09-18T12:04:09Z

PR Docs

PaddlePaddle/docs#6878

PR APIs

torch.nn.functional.lp_pool1d torch.nn.functional.lp_pool2d torch.nn.functional.threshold_ torch.nn.functional.feature_alpha_dropout torch.nn.functional.scaled_dot_product_attention torch.nn.LPPool1d torch.nn.LPPool2d torch.nn.Softmin torch.nn.AdaptiveLogSoftmaxWithLoss torch.nn.parameter.UninitializedParameter torch.nn.parameter.UninitializedBuffer torch.nn.CircularPad3d torch.nn.utils.parametrizations.weight_norm torch.optim.RAdam torch.optim.NAdam

目前问题：

lp_pool torch行为与Paddle不一致，当norm_type为inf时，应该与max_pool一致，而torch返回1。

未找到UninitializedParameter与UninitializedBuffer的映射

描述的问题的方式，从Pytorch的角度来描述。对Pytorch的某些具体功能，看Paddle的表现有什么差异，分析清楚差异是否是Paddle有功能Bug，或者功能缺陷。给出明确结论，不是简单的罗列问题。
这个不是都能有现成的映射，需要去分析出对应的替代实现方式，使得对网络计算无影响即可

Asthestarsfalll · 2024-09-19T01:27:35Z

PR Docs

PaddlePaddle/docs#6878

PR APIs

torch.nn.functional.lp_pool1d torch.nn.functional.lp_pool2d torch.nn.functional.threshold_ torch.nn.functional.feature_alpha_dropout torch.nn.functional.scaled_dot_product_attention torch.nn.LPPool1d torch.nn.LPPool2d torch.nn.Softmin torch.nn.AdaptiveLogSoftmaxWithLoss torch.nn.parameter.UninitializedParameter torch.nn.parameter.UninitializedBuffer torch.nn.CircularPad3d torch.nn.utils.parametrizations.weight_norm torch.optim.RAdam torch.optim.NAdam
目前问题：

lp_pool torch行为与Paddle不一致，当norm_type为inf时，应该与max_pool一致，而torch返回1。

未找到UninitializedParameter与UninitializedBuffer的映射

描述的问题的方式，从Pytorch的角度来描述。对Pytorch的某些具体功能，看Paddle的表现有什么差异，分析清楚差异是否是Paddle有功能Bug，或者功能缺陷。给出明确结论，不是简单的罗列问题。

这个不是都能有现成的映射，需要去分析出对应的替代实现方式，使得对网络计算无影响即可

根据PyTorch的文档，当norm_type为inf时，应该与max_pool一致，Paddle的实现在norm_type为inf时，会直接调用max_pool，而PyTorch直接计算，由于浮点数范围原因，会输出为1
了解

zhwesky2010 · 2024-09-19T08:04:06Z

@Asthestarsfalll 单测失败，请保证CI通过，CI未通过不予合入：

2024-09-19 10:46:30 FAILED tests/test_nn_LPPool1d.py::test_case_3 - AssertionError: API (torch.nn...
2024-09-19 10:46:30 FAILED tests/test_nn_LPPool2d.py::test_case_6 - AssertionError: API (torch.nn...
2024-09-19 10:46:30 FAILED tests/test_nn_Softmin.py::test_case_1 - AttributeError: module 'paddle...
2024-09-19 10:46:30 FAILED tests/test_nn_Softmin.py::test_case_2 - AttributeError: module 'paddle...
2024-09-19 10:46:30 FAILED tests/test_nn_functional_lp_pool1d.py::test_case_5 - AssertionError: A...
2024-09-19 10:46:30 FAILED tests/test_nn_functional_lp_pool2d.py::test_case_6 - AssertionError: A...
2024-09-19 10:46:30 ============ 6 failed, 8152 passed, 90 skipped in 188.53s (0:03:08) ============

zhwesky2010 · 2024-09-19T08:05:27Z

@Asthestarsfalll 第1点问题，写一个屏蔽版本的单测，然后在注释备注好原因

Asthestarsfalll · 2024-09-25T00:18:42Z

paconvert/api_matcher.py

+            class Softmin(paddle.nn.Softmax):
+                def forward(self, x):
+                    return super().forward(-x)
+            setattr(paddle.nn, 'Softmin', Softmin)


好像没有办法获取forward的输入，所以只能这样

zhwesky2010 · 2024-09-29T10:02:05Z

@Asthestarsfalll 现在CI未通过，需要修复

zhwesky2010 · 2024-10-08T10:43:28Z

@Asthestarsfalll CI未通过，请自查问题

zhwesky2010 · 2024-10-12T06:53:50Z

@Asthestarsfalll CI仍未通过，请自查问题

zhwesky2010 · 2024-10-17T04:21:40Z

=========================== short test summary info ============================
2024-10-10 18:06:37 FAILED tests/test_nn_Softmin.py::test_case_1 - AttributeError: module 'paddle...
2024-10-10 18:06:37 FAILED tests/test_nn_Softmin.py::test_case_2 - AttributeError: module 'paddle...
2024-10-10 18:06:37 =========== 2 failed, 8289 passed, 105 skipped in 250.09s (0:04:10) ============

CI未通过

Asthestarsfalll · 2024-10-22T01:36:29Z

tests/test_nn_Softmin.py

+
+from apibase import APIBase
+
+obj = APIBase("torch.nn.Softmin", is_aux_api=True)


@zhwesky2010 ci过不了应该是因为aux code，但是设置为True还是有问题

@zhwesky2010 ci过不了应该是因为aux code，但是设置为True还是有问题

你本地测试有问题吗？

如果本地测试就有问题，需要你自己逐行去调试。

如果本地测试没问题，只在CI上有问题，我们会帮忙排查或修复CI环境的问题。

zhwesky2010

这个你还是要查查Softmin哪里写的不对，逐行调试

zhwesky2010 · 2024-10-25T04:29:01Z

paconvert/api_matcher.py

@@ -3914,6 +3914,36 @@ def generate_code(self, kwargs):
        return GenericMatcher.generate_code(self, kwargs)


+class SoftminMatcher(SoftmaxMatcher):
+    def generate_code(self, kwargs):
+        self.paddle_api = "paddle.nn.Softmin"


这个在json里不能配吗

因为要修改forward，似乎不能通过json配吧

zhwesky2010 · 2024-10-25T04:29:55Z

paconvert/api_matcher.py

+            class Softmin(paddle.nn.Softmax):
+                def forward(self, x):
+                    return super().forward(-x)
+            setattr(paddle.nn, 'Softmin', Softmin)


这个就不用setattr paddle了，直接调用paddle_aux.Softmin吧，避免别人认为paddle也有paddle.nn.Softmin

已更新，本地测试截图：

本地第一次运行会出现同样的错误，但是第二次运行就正常了，测试了其他使用aux code的文件，也有同样的问题，比如Softmax

本地第一次运行会出现同样的错误，但是第二次运行就正常了，测试了其他使用aux code的文件，也有同样的问题，比如Softmax

@zhwesky2010

tests/test_optim_NAdam.py

tests/test_optim_RAdam.py

paconvert/api_mapping.json

zhwesky2010 · 2024-11-12T10:25:47Z

@Asthestarsfalll 这个需要修复一下CI

luotao1 · 2024-11-25T08:44:52Z

除去torch.nn.Softmin，其他的可以先提个PR，合入一版么？

zhwesky2010

合并下最新的吧，目前落后master的commit很多了
单测case务必全面，以下四种情况的测试case必须全部包含：

1. 传入所有参数且全部指定关键字
2. 传入所有参数且全部不指定关键字
3. 默认参数均不指定
4. 改变关键字顺序

paconvert/api_mapping.json

paconvert/api_matcher.py

tests/test_nn_Softmin.py

tests/test_nn_functional_threshold_.py

tests/test_optim_RAdam.py

zhwesky2010 · 2024-11-26T04:30:38Z

paconvert/api_matcher.py

+                if self._axis is None:
+                    return paddle.nn.functional.softmax(x, _get_softmax_dim(x.ndim))
+                return paddle.nn.functional.softmax(x, self._axis)
+            setattr(paddle.nn.Softmax, 'forward', forward)


那就不改paddle.nn.Softmax，直接把这些实现放到Softmin下面就行，遵循辅助函数最小原则，非必要不改动

zhwesky2010 · 2024-11-26T04:31:26Z

tests/test_nn_AdaptiveLogSoftmaxWithLoss.py

+        out, loss = asfm(input,target)
+        """
+    )
+    obj.run(pytorch_code, ["out", "loss"], check_value=False)


这个有随机性吗，为何关闭了check_value

这两个都有随机初始化的权重

tests/test_nn_utils_parametrizations_weight_norm.py

zhwesky2010 · 2024-11-26T13:24:51Z

paconvert/api_matcher.py

+                if self._axis is None:
+                    return paddle.nn.functional.softmax(x, _get_softmax_dim(x.ndim))
+                return paddle.nn.functional.softmax(x, self._axis)
+            setattr(paddle.nn.Softmax, 'forward', forward)


我的意思是：你不要去调整paddle.nn.Softmax，这个和paddle.nn.Softmax有什么关系呢，为什么非要去改动其他的api呢

【Hackathon 7th No.39】为 Paddle 代码转换工具新增 API 转换规则（第 6 组）

469a523

luotao1 mentioned this pull request Sep 18, 2024

【Hackathon 7th】开源贡献个人挑战赛 PaddlePaddle/Paddle#68244

Open

paddle-bot bot added the contributor External developers label Sep 18, 2024

luotao1 assigned luotao1 and zhwesky2010 Sep 19, 2024

Asthestarsfalll commented Sep 25, 2024

View reviewed changes

Asthestarsfalll and others added 2 commits October 8, 2024 19:03

Merge branch 'master' into 6

a1d6902

add skip

68fa44f

Asthestarsfalll commented Oct 22, 2024

View reviewed changes

zhwesky2010 reviewed Oct 25, 2024

View reviewed changes

Asthestarsfalll and others added 2 commits October 25, 2024 15:33

update Softmin

e590a24

Merge branch 'master' into 6

0b7b6b2

PaddlePaddle locked and limited conversation to collaborators Nov 5, 2024

PaddlePaddle unlocked this conversation Nov 5, 2024

zhwesky2010 reviewed Nov 5, 2024

View reviewed changes

tests/test_optim_NAdam.py Show resolved Hide resolved

tests/test_optim_RAdam.py Show resolved Hide resolved

paconvert/api_mapping.json Show resolved Hide resolved

paconvert/api_mapping.json Show resolved Hide resolved

PaddlePaddle locked and limited conversation to collaborators Nov 5, 2024

PaddlePaddle unlocked this conversation Nov 5, 2024

Asthestarsfalll added 2 commits November 7, 2024 10:47

update

ff99c66

update

340d69d

Merge branch 'PaddlePaddle:master' into 6

cbe055f

Asthestarsfalll added 2 commits November 25, 2024 19:01

fix

a6e5ed5

fix codestyle check

555dbd3

zhwesky2010 reviewed Nov 25, 2024

View reviewed changes

Asthestarsfalll and others added 2 commits November 26, 2024 01:12

fix

0f8d128

Merge branch 'PaddlePaddle:master' into 6

9a1e17e

zhwesky2010 reviewed Nov 26, 2024

View reviewed changes

fix

7eb8406

zhwesky2010 reviewed Nov 26, 2024

View reviewed changes

fix

158f924

zhwesky2010 approved these changes Nov 27, 2024

View reviewed changes

zhwesky2010 merged commit 3c73e2e into PaddlePaddle:master Nov 27, 2024
7 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

【Hackathon 7th No.39】为 Paddle 代码转换工具新增 API 转换规则（第 6 组） #477

【Hackathon 7th No.39】为 Paddle 代码转换工具新增 API 转换规则（第 6 组） #477

Asthestarsfalll commented Sep 18, 2024 •

edited by zhwesky2010

Loading

paddle-bot bot commented Sep 18, 2024

zhwesky2010 commented Sep 18, 2024 •

edited

Loading

PR Docs

PR APIs

Asthestarsfalll commented Sep 19, 2024

PR Docs

PR APIs

zhwesky2010 commented Sep 19, 2024 •

edited

Loading

zhwesky2010 commented Sep 19, 2024

Asthestarsfalll Sep 25, 2024

zhwesky2010 commented Sep 29, 2024

zhwesky2010 commented Oct 8, 2024

zhwesky2010 commented Oct 12, 2024

zhwesky2010 commented Oct 17, 2024

Asthestarsfalll Oct 22, 2024

zhwesky2010 Oct 25, 2024 •

edited

Loading

zhwesky2010 left a comment

zhwesky2010 Oct 25, 2024

Asthestarsfalll Oct 25, 2024

zhwesky2010 Oct 25, 2024

Asthestarsfalll Oct 25, 2024

Asthestarsfalll Oct 27, 2024

Asthestarsfalll Nov 14, 2024

zhwesky2010 commented Nov 12, 2024

luotao1 commented Nov 25, 2024

zhwesky2010 left a comment

zhwesky2010 Nov 26, 2024

zhwesky2010 Nov 26, 2024

Asthestarsfalll Nov 26, 2024

zhwesky2010 Nov 26, 2024

Asthestarsfalll Nov 26, 2024


		from apibase import APIBase

		obj = APIBase("torch.nn.Softmin", is_aux_api=True)

【Hackathon 7th No.39】为 Paddle 代码转换工具新增 API 转换规则（第 6 组） #477

【Hackathon 7th No.39】为 Paddle 代码转换工具新增 API 转换规则（第 6 组） #477

Conversation

Asthestarsfalll commented Sep 18, 2024 • edited by zhwesky2010 Loading

PR Docs

PR APIs

paddle-bot bot commented Sep 18, 2024

zhwesky2010 commented Sep 18, 2024 • edited Loading

PR Docs

PR APIs

Asthestarsfalll commented Sep 19, 2024

PR Docs

PR APIs

zhwesky2010 commented Sep 19, 2024 • edited Loading

zhwesky2010 commented Sep 19, 2024

Choose a reason for hiding this comment

zhwesky2010 commented Sep 29, 2024

zhwesky2010 commented Oct 8, 2024

zhwesky2010 commented Oct 12, 2024

zhwesky2010 commented Oct 17, 2024

Choose a reason for hiding this comment

zhwesky2010 Oct 25, 2024 • edited Loading

Choose a reason for hiding this comment

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

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

zhwesky2010 commented Nov 12, 2024

luotao1 commented Nov 25, 2024

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

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Asthestarsfalll commented Sep 18, 2024 •

edited by zhwesky2010

Loading

zhwesky2010 commented Sep 18, 2024 •

edited

Loading

zhwesky2010 commented Sep 19, 2024 •

edited

Loading

zhwesky2010 Oct 25, 2024 •

edited

Loading