【Hackathon 5th No.33】为 Paddle 新增 atleast_1d / atleast_2d / atleast_3d API

[megemini]

PR types

Others

PR changes

Docs

Description

【Hackathon 5th No.33】为 Paddle 新增 atleast_1d / atleast_2d / atleast_3d API

请评审！

[paddle-bot]

你的PR提交成功，感谢你对开源项目的贡献!
请检查PR提交格式和内容是否完备，具体请参考示例和模版。
Your PR has been submitted. Thanks for your contribution!
Please check its format and content. For this, you can refer to Template and Demo.

[luotao1]

数据类型支持：float32、float64、int32、int64。

数据类型上：float16, uint16, float32, float64, int8, int16, int32, int64, uint8, complex64, complex128, bfloat16，这些都能支持么？

[megemini]

这个我再确认一下，因为 manipulation.py 里面很多都是 float32、float64、int32、int64 类似的类型，所以这里是对齐此类方法 ～

后面确认后也会加到单测里面 ～

[megemini]

Update 20231012

• 增加 name 参数，参考 【Hackathon 5th No.32】为 Paddle 新增 tensor_split / hsplit / dsplit API #682 (comment) 的讨论
• 修改支持数据类型

简单测试了一下数据类型的支持情况：

In [32]: atleast_1d(float16, uint16, float32, float64, int8, int16, int32, int64, uint8, complex64, complex128, bfloat16)
    ...: 
Out[32]: 
[Tensor(shape=[1], dtype=float16, place=Place(cpu), stop_gradient=True,
        [0.30004883]),
 Tensor(shape=[1], dtype=bfloat16, place=Place(cpu), stop_gradient=True,
        [23.]),
 Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True,
        [3.]),
 Tensor(shape=[1], dtype=float64, place=Place(cpu), stop_gradient=True,
        [23.]),
 Tensor(shape=[1], dtype=int8, place=Place(cpu), stop_gradient=True,
        [2]),
 Tensor(shape=[1], dtype=int16, place=Place(cpu), stop_gradient=True,
        [2]),
 Tensor(shape=[1], dtype=int32, place=Place(cpu), stop_gradient=True,
        [2]),
 Tensor(shape=[1], dtype=int64, place=Place(cpu), stop_gradient=True,
        [2]),
 Tensor(shape=[1], dtype=uint8, place=Place(cpu), stop_gradient=True,
        [2]),
 Tensor(shape=[1], dtype=complex64, place=Place(cpu), stop_gradient=True,
        [(1+1j)]),
 Tensor(shape=[1], dtype=complex128, place=Place(cpu), stop_gradient=True,
        [(1+1j)]),
 Tensor(shape=[1], dtype=bfloat16, place=Place(cpu), stop_gradient=True,
        [0.29882812])]

In [33]: atleast_2d(float16, uint16, float32, float64, int8, int16, int32, int64, uint8, complex64, complex128, bfloat16)
    ...: 
Out[33]: 
[Tensor(shape=[1, 1], dtype=float16, place=Place(cpu), stop_gradient=True,
        [[0.30004883]]),
 Tensor(shape=[1, 1], dtype=bfloat16, place=Place(cpu), stop_gradient=True,
        [[23.]]),
 Tensor(shape=[1, 1], dtype=float32, place=Place(cpu), stop_gradient=True,
        [[3.]]),
 Tensor(shape=[1, 1], dtype=float64, place=Place(cpu), stop_gradient=True,
        [[23.]]),
 Tensor(shape=[1, 1], dtype=int8, place=Place(cpu), stop_gradient=True,
        [[2]]),
 Tensor(shape=[1, 1], dtype=int16, place=Place(cpu), stop_gradient=True,
        [[2]]),
 Tensor(shape=[1, 1], dtype=int32, place=Place(cpu), stop_gradient=True,
        [[2]]),
 Tensor(shape=[1, 1], dtype=int64, place=Place(cpu), stop_gradient=True,
        [[2]]),
 Tensor(shape=[1, 1], dtype=uint8, place=Place(cpu), stop_gradient=True,
        [[2]]),
 Tensor(shape=[1, 1], dtype=complex64, place=Place(cpu), stop_gradient=True,
        [[(1+1j)]]),
 Tensor(shape=[1, 1], dtype=complex128, place=Place(cpu), stop_gradient=True,
        [[(1+1j)]]),
 Tensor(shape=[1, 1], dtype=bfloat16, place=Place(cpu), stop_gradient=True,
        [[0.29882812]])]

In [34]: atleast_3d(float16, uint16, float32, float64, int8, int16, int32, int64, uint8, complex64, complex128, bfloat16)
    ...: 
Out[34]: 
[Tensor(shape=[1, 1, 1], dtype=float16, place=Place(cpu), stop_gradient=True,
        [[[0.30004883]]]),
 Tensor(shape=[1, 1, 1], dtype=bfloat16, place=Place(cpu), stop_gradient=True,
        [[[23.]]]),
 Tensor(shape=[1, 1, 1], dtype=float32, place=Place(cpu), stop_gradient=True,
        [[[3.]]]),
 Tensor(shape=[1, 1, 1], dtype=float64, place=Place(cpu), stop_gradient=True,
        [[[23.]]]),
 Tensor(shape=[1, 1, 1], dtype=int8, place=Place(cpu), stop_gradient=True,
        [[[2]]]),
 Tensor(shape=[1, 1, 1], dtype=int16, place=Place(cpu), stop_gradient=True,
        [[[2]]]),
 Tensor(shape=[1, 1, 1], dtype=int32, place=Place(cpu), stop_gradient=True,
        [[[2]]]),
 Tensor(shape=[1, 1, 1], dtype=int64, place=Place(cpu), stop_gradient=True,
        [[[2]]]),
 Tensor(shape=[1, 1, 1], dtype=uint8, place=Place(cpu), stop_gradient=True,
        [[[2]]]),
 Tensor(shape=[1, 1, 1], dtype=complex64, place=Place(cpu), stop_gradient=True,
        [[[(1+1j)]]]),
 Tensor(shape=[1, 1, 1], dtype=complex128, place=Place(cpu), stop_gradient=True,
        [[[(1+1j)]]]),
 Tensor(shape=[1, 1, 1], dtype=bfloat16, place=Place(cpu), stop_gradient=True,
        [[[0.29882812]]])]

除了 uint16 会默认转换为 bfloat16 之外，暂时未发现问题。由此更新文档，并增加单测需要测试的数据类型。

@luotao1 请评审～