【Hackathon 7th No.38】为Paddle代码转换工具新增API转换规则（第5组） #6885

inaomIIsfarell · 2024-09-21T01:47:13Z

【Hackathon 7th】开源贡献个人挑战赛 Paddle#68244

CLAassistant · 2024-09-21T01:47:18Z

All committers have signed the CLA.

zhwesky2010 · 2024-09-23T08:16:25Z

docs/guides/model_convert/convert_from_pytorch/api_difference/Tensor/torch.Tensor.positive.md

@@ -0,0 +1,11 @@
+## [功能缺失]torch.Tensor.positive


这个想办法组合实现吧，简单的功能，不用判定为功能缺失。只有必须新增API 的才被认定为功能缺失

zhwesky2010 · 2024-09-23T08:31:35Z

...ides/model_convert/convert_from_pytorch/api_difference/Tensor/torch.Tensor.scatter_reduce.md

+### [torch.Tensor.scatter_reduce](https://pytorch.org/docs/stable/generated/torch.Tensor.scatter_reduce.html#torch-tensor-scatter-reduce)
+
+```python
+Tensor.scatter_reduce(dim, index, src, reduce, *, include_self=True)


API签名要写全名：torch.Tensor.scatter_reduce

zhwesky2010 · 2024-09-23T08:34:56Z

docs/guides/model_convert/convert_from_pytorch/api_difference/torch/torch.can_cast.md

+torch.can_cast(from_, to)
+```
+
+判断类型的转换在 PyTorch 的[casting 规则](https://pytorch.org/docs/stable/tensor_attributes.html#type-promotion-doc)中是否被允许，暂未发现 Paddle 中有 api 能实现该功能。


这个有无办法组合实现？

简单的功能，不用判定为功能缺失。只有必须新增API 的才被认定为功能缺失

Paddle也是支持类型提升的 https://www.paddlepaddle.org.cn/documentation/docs/zh/guides/advanced/auto_type_promotion_cn.html

需要写一个方法实现逻辑功能，在api_matcher中通过generate_aux_code方法实现转写，文档的转写示例里直接把方法写出来吗？另外tensor和paddle的类型提升支持范围有差异。

需要写一个方法实现逻辑功能，在api_matcher中通过generate_aux_code方法实现转写，文档的转写示例里直接把方法写出来吗？另外tensor和paddle的类型提升支持范围有差异。

可以这样做，简单组合直接走辅助函数，然后辅助函数的逻辑也需要在映射文档中呈现

从源码看，pytorch的无符号整数的类型支持uint8，uint16，uint32，uint64，但paddle关于无符号整数的类型只支持uint8；

paddle.Tensor.dtype共支持12种数据类型，我个人想到实现的方式是通过字典存储各类型的转换关系，通过转换前后的类型参数直接索引（事实上pytorch也是这么实现的）。但是这样代码行数太多，写在映射文档种是否不太合适？

zhwesky2010 · 2024-09-23T08:36:20Z

docs/guides/model_convert/convert_from_pytorch/api_difference/torch/torch.concatenate.md

+### [paddle.concat](https://www.paddlepaddle.org.cn/documentation/docs/zh/develop/api/paddle/concat_cn.html)
+
+```python
+paddle.concat(x, axis=0, name=None)


这个就按API alias别名来处理吧，配置在 https://github.com/PaddlePaddle/PaConvert/blob/master/paconvert/api_alias_mapping.json ，无需新增映射文档。

zhwesky2010 · 2024-09-23T08:37:34Z

docs/guides/model_convert/convert_from_pytorch/api_difference/torch/torch.float_power.md

+torch.float_power(input, exponent, *, out=None)
+```
+
+Paddle 无此 API，需要组合实现。


你这个需要写两个转写示例，先写正常情况下如何组合实现，再写 out如何转写

zhwesky2010 · 2024-09-23T08:37:49Z

docs/guides/model_convert/convert_from_pytorch/api_difference/torch/torch.float_power.md

+
+```python
+# PyTorch 写法
+torch.float_pow(x, y, out)


API签名写错？

zhwesky2010 · 2024-09-23T08:39:32Z

docs/guides/model_convert/convert_from_pytorch/api_difference/torch/torch.positive.md

+torch.positive(input)
+```
+
+判断 `input` 是否是 bool 类型的 Tensor，如果是则抛出 RuntimeError 异常，否则返回 `input` 。


同上，简单的功能尽可能找方法去组合实现，只有必须要Paddle 新增API的才计做功能缺失

zhwesky2010 · 2024-09-25T04:14:05Z

docs/guides/model_convert/convert_from_pytorch/api_difference/torch/torch.can_cast.md

+    }
+    return can_cast_dict[from_][to]
+
+can_cast(x, y)


这个情况比较多，需要核对下，是不是与Torch结果一致的

我写了一个测试代码，把示例中 can_cast_dict 的 paddle 全部替换成 torch，和 torch.can_cast 的结果一致

import torch can_cast_dict = { torch.bfloat16: { torch.bfloat16: True, torch.float16: True, torch.float32: True, torch.float64: True, torch.complex64: True, torch.complex128: True, torch.uint8: False, torch.int8: False, torch.int16: False, torch.int32: False, torch.int64: False, torch.bool: False }, torch.float16: { torch.bfloat16: True, torch.float16: True, torch.float32: True, torch.float64: True, torch.complex64: True, torch.complex128: True, torch.uint8: False, torch.int8: False, torch.int16: False, torch.int32: False, torch.int64: False, torch.bool: False, }, torch.float32: { torch.bfloat16: True, torch.float16: True, torch.float32: True, torch.float64: True, torch.complex64: True, torch.complex128: True, torch.uint8: False, torch.int8: False, torch.int16: False, torch.int32: False, torch.int64: False, torch.bool: False, }, torch.float64: { torch.bfloat16: True, torch.float16: True, torch.float32: True, torch.float64: True, torch.complex64: True, torch.complex128: True, torch.uint8: False, torch.int8: False, torch.int16: False, torch.int32: False, torch.int64: False, torch.bool: False, }, torch.complex64: { torch.bfloat16: False, torch.float16: False, torch.float32: False, torch.float64: False, torch.complex64: True, torch.complex128: True, torch.uint8: False, torch.int8: False, torch.int16: False, torch.int32: False, torch.int64: False, torch.bool: False, }, torch.complex128: { torch.bfloat16: False, torch.float16: False, torch.float32: False, torch.float64: False, torch.complex64: True, torch.complex128: True, torch.uint8: False, torch.int8: False, torch.int16: False, torch.int32: False, torch.int64: False, torch.bool: False, }, torch.uint8: { torch.bfloat16: True, torch.float16: True, torch.float32: True, torch.float64: True, torch.complex64: True, torch.complex128: True, torch.uint8: True, torch.int8: True, torch.int16: True, torch.int32: True, torch.int64: True, torch.bool: False, }, torch.int8: { torch.bfloat16: True, torch.float16: True, torch.float32: True, torch.float64: True, torch.complex64: True, torch.complex128: True, torch.uint8: True, torch.int8: True, torch.int16: True, torch.int32: True, torch.int64: True, torch.bool: False, }, torch.int16: { torch.bfloat16: True, torch.float16: True, torch.float32: True, torch.float64: True, torch.complex64: True, torch.complex128: True, torch.uint8: True, torch.int8: True, torch.int16: True, torch.int32: True, torch.int64: True, torch.bool: False, }, torch.int32: { torch.bfloat16: True, torch.float16: True, torch.float32: True, torch.float64: True, torch.complex64: True, torch.complex128: True, torch.uint8: True, torch.int8: True, torch.int16: True, torch.int32: True, torch.int64: True, torch.bool: False, }, torch.int64: { torch.bfloat16: True, torch.float16: True, torch.float32: True, torch.float64: True, torch.complex64: True, torch.complex128: True, torch.uint8: True, torch.int8: True, torch.int16: True, torch.int32: True, torch.int64: True, torch.bool: False, }, torch.bool: { torch.bfloat16: True, torch.float16: True, torch.float32: True, torch.float64: True, torch.complex64: True, torch.complex128: True, torch.uint8: True, torch.int8: True, torch.int16: True, torch.int32: True, torch.int64: True, torch.bool: True, } } for _from_dtype in can_cast_dict.keys(): for _to_dtype in can_cast_dict[_from_dtype].keys(): assert torch.can_cast(_from_dtype, _to_dtype) == can_cast_dict[_from_dtype][_to_dtype], "can_cast error" print(f"_from_dtype={_from_dtype}, _to_dtype={_to_dtype}") print("can_cast test pass")

zhwesky2010 · 2024-09-25T04:15:50Z

docs/guides/model_convert/convert_from_pytorch/api_difference/torch/torch.float_power.md

+torch.float_power(x, y)
+
+# Paddle 写法
+paddle.pow(paddle.cast(x, paddle.float64), y)


paddle.pow(x.float64(), y)

是不是看起来简洁些

inaomIIsfarell · 2024-09-26T04:57:57Z

...ides/model_convert/convert_from_pytorch/api_difference/Tensor/torch.Tensor.scatter_reduce.md

+| dim          | axis         | 表示 scatter 的维度，仅参数名不一致。                        |
+| index        | indices      | 表示输入的索引张量，仅参数名不一致。                         |
+| src          | values       | 表示需要插入的值，仅参数名不一致。                           |
+| reduce       | reduce       | 表示插入 values 时的计算方式，参数默认值不一致。PyTorch 中该参数无默认值，需要输入，Paddle 中默认值为 `assign`，应设置为与 PyTorch 一致。其中 PyTorch 的 `sum` 对应 Paddle 中的 `add`，PyTorch 的 `prod` 对应 Paddle 中 `multiply`。 |


此处 paddle 的 api 文档应该有误

此处 paddle 的 api 文档应该有误

以代码为准就行

inaomIIsfarell · 2024-09-26T07:05:15Z

.../guides/model_convert/convert_from_pytorch/api_difference/Tensor/torch.Tensor.float_power.md

这个api不是我负责的，我记错了把这个也改了。。。sorry
但是这么写应该也没问题

zhwesky2010 · 2024-09-27T08:05:44Z

docs/guides/model_convert/convert_from_pytorch/api_difference/torch/torch.isneginf.md

+torch.isneginf(a, out=b)
+
+# Paddle 写法
+paddle.assign(paddle.isneginf(x), y)


不要改参数名，上下对应，这样读者容易比对差异

zhwesky2010 · 2024-09-27T08:06:24Z

docs/guides/model_convert/convert_from_pytorch/api_difference/torch/torch.isposinf.md

+torch.isposinf(a, out=b)
+
+# Paddle 写法
+paddle.assign(paddle.isposinf(x), y)


不要改参数名，上下对应，这样读者容易比对差异

zhwesky2010 · 2024-09-29T08:58:22Z

@inaomIIsfarell CI还没通过：

2024-09-27 18:50:31 Exception: Unexpected End State at 9 in parsing file: /FluidDoc/docs/guides/model_convert/convert_from_pytorch/tools/../api_difference/torch/torch.isin.md, current meta: {'source_file': '/FluidDoc/docs/guides/model_convert/convert_from_pytorch/tools/../api_difference/torch/torch.isin.md', 'src_api': 'torch.isin', 'mapping_type': '?????????????????????', 'src_api_url': 'https://pytorch.org/docs/stable/generated/torch.isin.html#torch.isin', 'src_signature': {'api_name': 'torch.isin', 'args': [{'arg_name': 'elements', 'arg_default': None}, {'arg_name': 'test_elements', 'arg_default': None}, {'arg_name': '*', 'arg_default': None}, {'arg_name': 'assume_unique', 'arg_default': 'False'}, {'arg_name': 'invert', 'arg_default': 'False'}]}, 'dst_api': 'paddle.isin', 'dst_api_url': 'https://www.paddlepaddle.org.cn/documentation/docs/zh/develop/api/paddle/isin_cn.html', 'dst_signature': {'api_name': 'paddle.isin', 'args': [{'arg_name': 'x', 'arg_default': None}, {'arg_name': 'test_x', 'arg_default': None}, {'arg_name': 'assume_unique', 'arg_default': 'False'}, {'arg_name': 'invert', 'arg_default': 'False'}, {'arg_name': 'name', 'arg_default': 'None'}]}}
2024-09-27 18:50:31 Error: API mapping generate script failed, please check changes in /FluidDoc/docs/guides/model_convert/convert_from_pytorch

inaomIIsfarell · 2024-09-30T03:45:13Z

@inaomIIsfarell CI还没通过：

2024-09-27 18:50:31 Exception: Unexpected End State at 9 in parsing file: /FluidDoc/docs/guides/model_convert/convert_from_pytorch/tools/../api_difference/torch/torch.isin.md, current meta: {'source_file': '/FluidDoc/docs/guides/model_convert/convert_from_pytorch/tools/../api_difference/torch/torch.isin.md', 'src_api': 'torch.isin', 'mapping_type': '?????????????????????', 'src_api_url': 'https://pytorch.org/docs/stable/generated/torch.isin.html#torch.isin', 'src_signature': {'api_name': 'torch.isin', 'args': [{'arg_name': 'elements', 'arg_default': None}, {'arg_name': 'test_elements', 'arg_default': None}, {'arg_name': '*', 'arg_default': None}, {'arg_name': 'assume_unique', 'arg_default': 'False'}, {'arg_name': 'invert', 'arg_default': 'False'}]}, 'dst_api': 'paddle.isin', 'dst_api_url': 'https://www.paddlepaddle.org.cn/documentation/docs/zh/develop/api/paddle/isin_cn.html', 'dst_signature': {'api_name': 'paddle.isin', 'args': [{'arg_name': 'x', 'arg_default': None}, {'arg_name': 'test_x', 'arg_default': None}, {'arg_name': 'assume_unique', 'arg_default': 'False'}, {'arg_name': 'invert', 'arg_default': 'False'}, {'arg_name': 'name', 'arg_default': 'None'}]}}
2024-09-27 18:50:31 Error: API mapping generate script failed, please check changes in /FluidDoc/docs/guides/model_convert/convert_from_pytorch

这个不懂怎么解决诶。。

inaomIIsfarell added 2 commits September 20, 2024 23:06

【Hackathon 7th No.38】为 Paddle 代码转换工具新增 API 转换规则（第5组）

a9373ff

Merge branch 'develop' of https://github.com/PaddlePaddle/docs into hkt1

84fb9f7

paddle-bot bot added the contributor label Sep 21, 2024

fix docs

17e10a0

luotao1 added the PaddlePaddle Hackathon 飞桨黑客松活动issue与PR label Sep 23, 2024

luotao1 assigned luotao1, zhwesky2010 and sunzhongkai588 Sep 23, 2024

luotao1 mentioned this pull request Sep 23, 2024

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

Open

zhwesky2010 reviewed Sep 23, 2024

View reviewed changes

inaomIIsfarell added 4 commits September 25, 2024 10:42

fix

7697816

Merge branch 'develop' of https://github.com/PaddlePaddle/docs into hkt1

8c0edfb

fix

940ec07

update_param_exp

1a76e87

zhwesky2010 reviewed Sep 25, 2024

View reviewed changes

fix

571d1e8

inaomIIsfarell commented Sep 26, 2024

View reviewed changes

inaomIIsfarell mentioned this pull request Sep 26, 2024

【Hackathon 7th No.38】为 Paddle 代码转换工具新增 API 转换规则（第5组） PaddlePaddle/PaConvert#486

Closed

inaomIIsfarell commented Sep 26, 2024

View reviewed changes

inaomIIsfarell mentioned this pull request Sep 26, 2024

【Hackathon 7th No.38】为 Paddle 代码转换工具新增 API 转换规则（第5组） PaddlePaddle/PaConvert#487

Open

zhwesky2010 reviewed Sep 27, 2024

View reviewed changes

inaomIIsfarell added 2 commits September 27, 2024 18:47

fix

23bb1bd

Merge branch 'develop' of https://github.com/PaddlePaddle/docs into hkt1

de8ac23

inaomIIsfarell added 2 commits September 30, 2024 10:55

fix

eacf377

Merge branch 'develop' of https://github.com/PaddlePaddle/docs into hkt1

b66bfd3

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

【Hackathon 7th No.38】为Paddle代码转换工具新增API转换规则（第5组） #6885

【Hackathon 7th No.38】为Paddle代码转换工具新增API转换规则（第5组） #6885

inaomIIsfarell commented Sep 21, 2024

CLAassistant commented Sep 21, 2024 •

edited

Loading

zhwesky2010 Sep 23, 2024

zhwesky2010 Sep 23, 2024

zhwesky2010 Sep 23, 2024

inaomIIsfarell Sep 24, 2024

zhwesky2010 Sep 24, 2024 •

edited

Loading

inaomIIsfarell Sep 24, 2024 •

edited

Loading

zhwesky2010 Sep 23, 2024

zhwesky2010 Sep 23, 2024

zhwesky2010 Sep 23, 2024

zhwesky2010 Sep 23, 2024

zhwesky2010 Sep 25, 2024

inaomIIsfarell Sep 26, 2024

zhwesky2010 Sep 25, 2024

inaomIIsfarell Sep 26, 2024

zhwesky2010 Sep 27, 2024

inaomIIsfarell Sep 26, 2024

zhwesky2010 Sep 27, 2024

zhwesky2010 Sep 27, 2024

zhwesky2010 commented Sep 29, 2024 •

edited

Loading

inaomIIsfarell commented Sep 30, 2024

【Hackathon 7th No.38】为Paddle代码转换工具新增API转换规则（第5组） #6885

Are you sure you want to change the base?

【Hackathon 7th No.38】为Paddle代码转换工具新增API转换规则（第5组） #6885

Conversation

inaomIIsfarell commented Sep 21, 2024

CLAassistant commented Sep 21, 2024 • 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

zhwesky2010 Sep 24, 2024 • edited Loading

Choose a reason for hiding this comment

inaomIIsfarell Sep 24, 2024 • 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

zhwesky2010 commented Sep 29, 2024 • edited Loading

inaomIIsfarell commented Sep 30, 2024

CLAassistant commented Sep 21, 2024 •

edited

Loading

zhwesky2010 Sep 24, 2024 •

edited

Loading

inaomIIsfarell Sep 24, 2024 •

edited

Loading

zhwesky2010 commented Sep 29, 2024 •

edited

Loading