【Hackathon 7th No.42】NO.42 为 Paddle 代码转换工具新增 API 转换规则（第 9 组）-part #6893

decade-afk · 2024-09-26T08:07:38Z

No description provided.

zhwesky2010

文档需严格，同步提交PaConvert的Matcher和单测吧，用实际代码校验一遍

zhwesky2010 · 2024-09-27T08:23:45Z

.../model_convert/convert_from_pytorch/api_difference/Tensor/torch.Tensor.bitwise_left_shift.md

+### [paddle.bitwise_left_shift](https://www.paddlepaddle.org.cn/documentation/docs/zh/api/paddle/bitwise_left_shift_cn.html#bitwise-left-shift)
+
+```python
+paddle.bitwise_left_shift(x, y, is_arithmetic=True, out=None, name=None)


为啥不对应到 paddle.Tensor.bitwise_left_shift？

类API尽可能对应类API，除非你确实没有再用其他替代

因为真的没有

你这装的是什么版本的

麻烦import最新版本的paddle来验证

zhwesky2010 · 2024-09-27T08:23:49Z

...model_convert/convert_from_pytorch/api_difference/Tensor/torch.Tensor.bitwise_left_shift_.md

+### [paddle.bitwise_left_shift_](https://www.paddlepaddle.org.cn/documentation/docs/zh/develop/api/paddle/bitwise_left_shift__cn.html#bitwise-left-shift)
+
+```python
+paddle.bitwise_left_shift_(x, y, is_arithmetic=True, out=None, name=None)


zhwesky2010 · 2024-09-27T08:23:55Z

...model_convert/convert_from_pytorch/api_difference/Tensor/torch.Tensor.bitwise_right_shift.md

+### [paddle.bitwise_right_shift](https://www.paddlepaddle.org.cn/documentation/docs/zh/api/paddle/bitwise_right_shift_cn.html#bitwise-right-shift)
+
+```python
+paddle.bitwise_right_shift(x, y, is_arithmetic=True, out=None, name=None)


zhwesky2010 · 2024-09-27T08:23:59Z

...odel_convert/convert_from_pytorch/api_difference/Tensor/torch.Tensor.bitwise_right_shift_.md

+### [paddle.bitwise_right_shift_](https://www.paddlepaddle.org.cn/documentation/docs/zh/develop/api/paddle/bitwise_right_shift__cn.html#bitwise-right-shift)
+
+```python
+paddle.bitwise_right_shift_(x, y, is_arithmetic=True, out=None, name=None)


zhwesky2010 · 2024-09-27T08:24:16Z

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

+### [torch.Tensor.col_indices](https://pytorch.org/docs/stable/generated/torch.Tensor.col_indices.html)
+
+```python
+torch.Tensor.crow_indices()


你这个API签名对吗？

复制快了，忘改了，抱歉

zhwesky2010 · 2024-09-27T08:25:08Z

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

+torch.Tensor.dim_order()
+```
+
+获取张量在内存中的物理布局，PaddlePaddle 的 Tensor 默认是 contiguous 的, 因此可直接返回一个从0到 Tensor 的维度长度的列表即可。


直接删除了，合理吗？

上面写直接删除，这里写的是组合替代实现，但又不给出组合的示例

这个文档很不合格

我改成组合替代吧，这个抱歉，因为写的时候参考了is_continous的文档

zhwesky2010 · 2024-09-27T08:25:53Z

...uides/model_convert/convert_from_pytorch/api_difference/Tensor/torch.Tensor.is_sparse_csr.md

+### 转写示例
+
+```python
+# torch 版本可以直接访问属性


转写示例写简单点，核心一行就行，不用各种print、准备随机变量这些

zhwesky2010 · 2024-09-27T08:26:17Z

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

+### [paddle.Tensor.element_size](https://www.paddlepaddle.org.cn/documentation/docs/zh/develop/api/paddle/Tensor_cn.html#element-size)
+
+```python
+paddle.Tensor.element_size()


这两个数值验证了能对上吗

验证倒是还没有

zhwesky2010 · 2024-09-27T08:26:50Z

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

+
+```python
+# torch 版本
+x = torch.rand(3)


示例简洁，而且看起来有bug

这个我改一下

zhwesky2010 · 2024-09-27T08:27:23Z

...uides/model_convert/convert_from_pytorch/api_difference/Tensor/torch.Tensor.to_sparse_coo.md

+
+| PyTorch | PaddlePaddle | 备注                                                                                      |
+| ------- | ------------ | ----------------------------------------------------------------------------------------- |
+| -       | sparse_dim   | 在新的稀疏张量中包含的稀疏维度的数量，pytorch中无此参数，paddle令其为tensor维度长度即可。 |


paddle保持默认，就可以吧

sparse_dim他并没有默认值

没有tensor维度长度这个说法，改成 输入维度

zhwesky2010 · 2024-09-29T09:28:35Z

麻烦用最新版本的paddle来验证

luotao1 · 2024-10-08T08:16:37Z

已经弄好了，代码也跑通了

图片没有上传成功

decade-afk · 2024-10-08T08:20:10Z

麻烦用最新版本的paddle来验证

已经弄好了的

decade-afk · 2024-10-08T08:20:26Z

已经弄好了，代码也跑通了

图片没有上传成功

已经好了

zhwesky2010

提交reiview前先自查问题

zhwesky2010 · 2024-10-08T10:36:44Z

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

+y = a.nbytes
+
+# Paddle 写法
+y = paddle.numel(a) * a.element_size()


这个返回的是一个Tensor吧，应该需要返回一个int

zhwesky2010 · 2024-10-08T10:38:02Z

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

+### [torch.Tensor.stride](https://pytorch.org/docs/stable/generated/torch.Tensor.stride.html#torch-tensor-stride)
+
+```python
+torch.Tensor.stride(dim)


这个备注下 dim=None

zhwesky2010 · 2024-10-08T10:38:42Z

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

+x.stride(dim)
+
+# Paddle 版本
+x.get_strides() if dim is None else x.get_strides()[dim]


直接分行写吧，文档更注重可读性

if dim:

else:

zhwesky2010 · 2024-10-08T10:39:03Z

...uides/model_convert/convert_from_pytorch/api_difference/Tensor/torch.Tensor.to_sparse_coo.md

+
+| PyTorch | PaddlePaddle | 备注                                                                                      |
+| ------- | ------------ | ----------------------------------------------------------------------------------------- |
+| -       | sparse_dim   | 在新的稀疏张量中包含的稀疏维度的数量，pytorch 中无此参数，paddle 令其为 tensor 输入维度长度即可。 |


tensor输入维度

没有维度长度的说法，维度就是一个标量，标量没有长度。

好的好的，已经都改了的

decade-afk · 2024-10-12T02:26:25Z

doc和 PaddlePaddle/PaConvert#490 已经全都跑通了，请review @zhwesky2010

zhwesky2010 · 2024-10-16T04:16:19Z

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

+y = a.nbytes
+
+# Paddle 写法
+y = int(paddle.numel(a) * a.element_size())


为啥不用 a.size * a.element_size()，Tensor转来转去效率不更低吗

zhwesky2010 · 2024-10-16T04:16:59Z

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

+### [torch.Tensor.stride](https://pytorch.org/docs/stable/generated/torch.Tensor.stride.html#torch-tensor-stride)
+
+```python
+torch.Tensor.stride(dim)


torch.Tensor.stride(dim=None)

zhwesky2010 · 2024-10-16T04:20:22Z

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

+
+| PyTorch | PaddlePaddle | 备注                                                            |
+| ------- | ------------ | --------------------------------------------------------------- |
+| dim     | -            | 返回指定维度的步长，默认返回全部步长，paddle 不支持，需要转写。 |


返回指定维度的步长，Pytorch为可选值，默认返回全部步长，此时无需转写，当有输入值时，需要转写。

zhwesky2010

LGTM

【Hackathon 7th No.42】NO.42 为 Paddle 代码转换工具新增 API 转换规则（第 9 组）

6f2c565

luotao1 mentioned this pull request Sep 26, 2024

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

Open

paddle-bot bot added the contributor label Sep 26, 2024

luotao1 assigned luotao1, sunzhongkai588 and zhwesky2010 Sep 27, 2024

luotao1 added the PaddlePaddle Hackathon 飞桨黑客松活动issue与PR label Sep 27, 2024

zhwesky2010 reviewed Sep 27, 2024

View reviewed changes

decade-afk mentioned this pull request Sep 28, 2024

【Hackathon 7th No.42】NO.42 为 Paddle 代码转换工具新增 API 转换规则（第 9 组） PaddlePaddle/PaConvert#490

Merged

Merge branch 'PaddlePaddle:develop' into itemsize

99a57c8

decade-afk closed this Oct 2, 2024

decade-afk reopened this Oct 2, 2024

update

a4f97e4

decade-afk force-pushed the itemsize branch from c6db54e to a4f97e4 Compare October 2, 2024 07:50

decade-afk added 5 commits October 2, 2024 15:55

update

d8700c8

update

6ba1da0

update

c2b8407

update

a8d64f2

update

885d33c

zhwesky2010 reviewed Oct 8, 2024

View reviewed changes

decade-afk added 3 commits October 9, 2024 18:46

update

1e41589

update

5e96059

update

2e3825c

zhwesky2010 reviewed Oct 16, 2024

View reviewed changes

update

65519ab

update

460c272

zhwesky2010 approved these changes Oct 16, 2024

View reviewed changes

zhwesky2010 merged commit 5cec675 into PaddlePaddle:develop Oct 16, 2024
2 checks passed

luotao1 changed the title ~~【Hackathon 7th No.42】NO.42 为 Paddle 代码转换工具新增 API 转换规则（第 9 组）~~ 【Hackathon 7th No.42】NO.42 为 Paddle 代码转换工具新增 API 转换规则（第 9 组）-part Oct 16, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

【Hackathon 7th No.42】NO.42 为 Paddle 代码转换工具新增 API 转换规则（第 9 组）-part #6893

【Hackathon 7th No.42】NO.42 为 Paddle 代码转换工具新增 API 转换规则（第 9 组）-part #6893

decade-afk commented Sep 26, 2024

zhwesky2010 left a comment •

edited

Loading

zhwesky2010 Sep 27, 2024

decade-afk Sep 27, 2024

zhwesky2010 Sep 29, 2024 •

edited

Loading

zhwesky2010 Sep 27, 2024

zhwesky2010 Sep 27, 2024

zhwesky2010 Sep 27, 2024

zhwesky2010 Sep 27, 2024

decade-afk Sep 27, 2024

zhwesky2010 Sep 27, 2024

decade-afk Sep 27, 2024

zhwesky2010 Sep 27, 2024

zhwesky2010 Sep 27, 2024

decade-afk Sep 27, 2024

zhwesky2010 Sep 27, 2024

decade-afk Sep 27, 2024

zhwesky2010 Sep 27, 2024

decade-afk Sep 27, 2024

zhwesky2010 Sep 29, 2024

zhwesky2010 commented Sep 29, 2024 •

edited

Loading

luotao1 commented Oct 8, 2024

decade-afk commented Oct 8, 2024

decade-afk commented Oct 8, 2024

zhwesky2010 left a comment

zhwesky2010 Oct 8, 2024

zhwesky2010 Oct 8, 2024

zhwesky2010 Oct 8, 2024

zhwesky2010 Oct 8, 2024

decade-afk Oct 9, 2024 •

edited

Loading

decade-afk commented Oct 12, 2024

zhwesky2010 Oct 16, 2024

zhwesky2010 Oct 16, 2024

zhwesky2010 Oct 16, 2024

zhwesky2010 left a comment

【Hackathon 7th No.42】NO.42 为 Paddle 代码转换工具新增 API 转换规则（第 9 组）-part #6893

【Hackathon 7th No.42】NO.42 为 Paddle 代码转换工具新增 API 转换规则（第 9 组）-part #6893

Conversation

decade-afk commented Sep 26, 2024

zhwesky2010 left a comment • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

zhwesky2010 Sep 29, 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

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

luotao1 commented Oct 8, 2024

decade-afk commented Oct 8, 2024

decade-afk commented Oct 8, 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

decade-afk Oct 9, 2024 • edited Loading

Choose a reason for hiding this comment

decade-afk commented Oct 12, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

zhwesky2010 left a comment

Choose a reason for hiding this comment

zhwesky2010 left a comment •

edited

Loading

zhwesky2010 Sep 29, 2024 •

edited

Loading

zhwesky2010 commented Sep 29, 2024 •

edited

Loading

decade-afk Oct 9, 2024 •

edited

Loading