[Docathon][Add API Legend No.55、22] add paddle.index_fill and paddle.shard_index API Paddle

[xu8117]

PR Category

PR Types

Description

[paddle-bot]

你的PR提交成功，感谢你对开源项目的贡献!
请关注后续CI自动化测试结果，详情请参考Paddle-CI手册。
Your PR has been submitted. Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

[sunzhongkai588]

• 图例部分不要写在代码示例的下面，和中文文档保持同步，写在api介绍的下方
• .. image:: .... 和上方文字空一行，否则无法渲染

[xu8117]

ci过不了，是否要重新提交一个pr

[sunzhongkai588]

图例解释可以精简一些，如

Suggested change

	A two-dimensional tensor with a shape of [3, 3], through the index_fill operation, when axis = 0, the index tensor index is [0, 2], and the fill value value = -1,
	all elements in the first and third rows are filled with -1, thereby obtaining a new tensor with a shape still [3, 3] but some elements changed. A two-dimensional
	tensor with the shape [3, 3], by means of the index_fill operation, when axis = 0, the index tensor index is [0, 2], and the fill value value = -1,
	all the elements in the first and third rows are filled with -1, thus getting a new tensor with the same shape [3, 3] but some elements changed.
	As shown below, a 2D tensor with shape [3, 3] is updated using the ``index_fill`` operation. With ``axis=0``, ``index=[0, 2]``, and ``value=-1``, the elements in the first and third rows are replaced with ``-1``.
	The resulting tensor retains the same shape [3, 3], but with updated values.

[sunzhongkai588]

Suggested change

	A two-dimensional tensor with a shape of [2, 1], through the shard_index operation, when index_num = 20, nshards = 2, and shard_id = 0, calculates
	the shard size shard_size = (20 + 2 - 1) // 2 = 10. For the element values in label, such as 16, because 16 >= 10 (not within the [0, 10) interval
	of shard 0), its value is set to ignore_value (i.e., -1) according to the formula; while the element value 1, since 0 <= 1 < 10 (within shard 0),
	the updated value is 1 - 0 * 10 = 1, and finally a new tensor with the same shape [2, 1] but some elements changed (such as shard_label) is obtained.
	As shown below, a 2D tensor with shape [2, 1] is updated using the ``shard_index`` operation. With ``index_num=20``, ``nshards=2``, and ``shard_id=0``, the shard size is calculated as ``shard_size = (20 + 2 - 1) // 2 = 10``.
	For each element in label: If the value is within the shard range ``[0, 10)``, it is adjusted to its offset in the range. For example, ``1`` is updated to ``1 - 0 * 10 = 1``; Otherwise, it is set to the ``ignore_value`` (default ``-1``). For example, ``16`` is outside the range ``[0, 10)`` and becomes ``-1``.

[paddle-ci-bot]

Sorry to inform you that dce184d's CIs have passed for more than 7 days. To prevent PR conflicts, you need to re-run all CIs manually.

[luotao1]

不需要重新提交，把 PR-CI-Codestyle-Check 流水线过了就行