[SWE-Paddle] Add code for PaddlePaddle__Paddle-58323

swe-paddle/tasks/PaddlePaddle__Paddle-58323/README.md

@@ -0,0 +1,37 @@
+# PaddlePaddle__Paddle-58323
+
+This directory converts Paddle PR #58323 into a SWE-Paddle community task candidate.
+
+## Source
+
+| Field | Value |
+| --- | --- |
+| Repo | `PaddlePaddle/Paddle` |
+| PR | [58323](https://github.com/PaddlePaddle/Paddle/pull/58323) |
+| PR title | 【Hackathon 5th No.33】为 Paddle 新增 atleast_1d / atleast_2d / atleast_3d API -part |
+| Base commit | `431a0d53bd1578385ebcd3021f08d88ed6f75c70` |
+| Merged at | `2023-11-16T12:29:50Z` |
+| Hackathon | `5th` task `33` |
+| Task type | `feature_enhancement` |
+| Resource | CPU |
+
+## Summary
+
+Add `atleast_1d`, `atleast_2d`, `atleast_3d` APIs for Paddle, which convert scalar or low-dimensional inputs to tensors with at least the requested number of dimensions, preserving higher-dimensional inputs unchanged.
+
+## Files
+
+- `proposal.md`: candidate proposal for maintainer triage.
+- `instruction.md`: self-contained problem statement for the coding agent.
+- `solution/code.patch`: gold patch from the merged PR.
+- `tests/test.patch`: test patch exposing the target behavior.
+- `tests/test.sh`: minimal target test command.
+- `environment/README.md`: environment notes for reproduction.
+
+## Verification
+
+```bash
+bash tests/test.sh
+```
+
+Expected behavior: applying `tests/test.patch` to `base_commit` should fail on the target behavior; applying both `tests/test.patch` and `solution/code.patch` should pass the target tests.

swe-paddle/tasks/PaddlePaddle__Paddle-58323/environment/README.md

@@ -0,0 +1,26 @@
+# Environment Notes
+
+This candidate is part of the SWE-Paddle task set.
+
+## Expected Environment
+
+- Repository: `PaddlePaddle/Paddle`
+- Base commit: `431a0d53bd1578385ebcd3021f08d88ed6f75c70`
+- Resource: CPU
+- Build path: Paddle source checkout at the base commit. This task only involves Python-side changes (tensor manipulation APIs built on existing `reshape`/`unsqueeze`), so a source build with Python tests is sufficient.
+
+## Run Order
+
+1. Check out `PaddlePaddle/Paddle` at the base commit.
+2. Apply `tests/test.patch`.
+3. Run `bash tests/test.sh`; the target behavior should fail before the fix.
+4. Apply `solution/code.patch`.
+5. Run `bash tests/test.sh` again; the target behavior should pass after the gold patch.
+
+## Minimal Test Command
+
+```bash
+bash tests/test.sh
+```
+
+The verifier is responsible for deriving stable F2P and P2P node IDs from repeated runs.

swe-paddle/tasks/PaddlePaddle__Paddle-58323/instruction.md

@@ -0,0 +1,35 @@
+# 新增 atleast_1d / atleast_2d / atleast_3d API
+
+## 详细描述
+
+为 Paddle 新增 `atleast_1d`、`atleast_2d`、`atleast_3d` 三个张量操作 API。它们将标量或低维输入转换为至少具有指定维度的张量,高维输入保持不变。对标量输入(0-D)会先转换为 1-D 张量,不足目标维度时通过 reshape / unsqueeze 补维。
+
+要求支持:
+
+- `paddle.atleast_1d(*inputs, name=None)`:输出至少 1-D
+  - 0-D 输入 reshape 为 `[1]`
+  - 1-D 及以上输入原样返回
+- `paddle.atleast_2d(*inputs, name=None)`:输出至少 2-D
+  - 0-D 输入 reshape 为 `[1, 1]`
+  - 1-D 输入 unsqueeze 到 `[1, N]`
+  - 2-D 及以上输入原样返回
+- `paddle.atleast_3d(*inputs, name=None)`:输出至少 3-D
+  - 0-D、1-D、2-D 输入分别补维到 3-D
+  - 3-D 及以上输入原样返回
+- 支持可变参数:单输入返回单个 Tensor,多输入返回 Tensor 列表
+- 输入可为 Tensor 或可被 `paddle.to_tensor` 转换的标量
+- 在 `paddle` 顶层命名空间、`paddle.tensor` 命名空间及 tensor 方法中导出
+- 支持的数据类型:float16, float32, float64, int16, int32, int64, int8, uint8, complex64, complex128, bfloat16, bool
+
+## 验收说明
+
+- `atleast_1d` / `atleast_2d` / `atleast_3d` 前向行为与 NumPy 对应 API(`np.atleast_1d/2d/3d`)一致
+- 单输入返回 Tensor,多输入返回 Tensor 列表
+- 高维输入保持不变,低维输入正确补维
+- 静态图与动态图下均可用
+
+## Acceptance Criteria
+
+- The behavior described above should be fixed.
+- Existing valid behavior should remain unchanged.
+- Do not satisfy the task by deleting tests, weakening assertions, or bypassing validation broadly.

swe-paddle/tasks/PaddlePaddle__Paddle-58323/proposal.md

@@ -0,0 +1,54 @@
+# Task Proposal: PaddlePaddle__Paddle-58323
+
+## 1. 来源信息
+
+- Instance ID：`PaddlePaddle__Paddle-58323`
+- PR 链接：https://github.com/PaddlePaddle/Paddle/pull/58323
+- PR 标题：`【Hackathon 5th No.33】为 Paddle 新增 atleast_1d / atleast_2d / atleast_3d API -part`
+- `base_commit`：`431a0d53bd1578385ebcd3021f08d88ed6f75c70`
+- merge commit：`557499bd9b511515309bcf8665b215ef7f016e71`
+- merged 时间：`2023-11-16T12:29:50Z`
+- 后续联系人：megemini
+
+## 2. 问题一句话
+
+为 Paddle 新增 `atleast_1d` / `atleast_2d` / `atleast_3d` API，将标量或低维输入转换为至少具有指定维度的张量。
+
+## 3. 为什么适合作为 SWE-Paddle 样本
+
+- **真实性**：该任务来自已合入的 Paddle Hackathon 5th 框架开发 PR，关联 RFC（community#679），不是合成任务。
+- **代表性**：它覆盖 Python 层张量操作 API 的完整落地链路：`paddle` 顶层命名空间与 `paddle.tensor` 命名空间导出、tensor 方法注册、可变参数处理（单输入返回 Tensor / 多输入返回列表）以及基于既有 reshape/unsqueeze 的实现。
+- **边界清楚**：目标行为集中在三个 API 的补维语义（0-D/1-D/2-D 输入分别如何补维），测试补丁可直接暴露目标行为。
+- **非平凡性**：需要正确处理可变参数、标量输入转换、不同维度的分支逻辑，并保证与 NumPy 语义一致；非纯配置修改。
+
+## 4. 任务类型和标签
+
+- 任务类型：`feature_enhancement`
+- 执行后端：`cpu`
+- 设备范围：`cpu`
+- 模块标签：`[python_api, tensor, manipulation]`
+
+## 5. 验证思路
+
+- 目标测试命令：`bash tests/test.sh`
+- 目标测试文件：`test/legacy_test/test_atleast_nd.py`
+- 修复前预期：在 `base_commit` 上应用 `tests/test.patch` 后，`test_atleast_nd.py` 中 `atleast_1d/2d/3d` 相关测试应 fail（API 不存在）。
+- 修复后预期：继续应用 `solution/code.patch` 后，目标测试应 pass。
+- P2P 候选：`test_atleast_nd.py` 为 PR 新增文件，无存量测试；可从同模块存量 manipulation 测试中选取回归护栏，由 verifier 自动抽取稳定 nodeid。
+
+## 6. 环境与资源
+
+- 资源需求：CPU
+- Paddle 来源：`PaddlePaddle/Paddle` source checkout at `base_commit`
+- 是否能提供 Docker：暂无，建议后续补充 source-build Dockerfile
+- patch 类型：纯 Python（tensor manipulation API，基于既有 reshape/unsqueeze 实现）
+- 环境建议：该样本仅涉及 Python 层改动，source build 后运行 Python 测试即可
+- 最小测试命令：`bash tests/test.sh`
+- 是否有 oracle 日志：由 SWE-Paddle verifier 结果另行维护
+
+## 7. 风险自查
+
+- 泄露风险：正式 `instruction.md` 应描述目标行为和验收标准，不直接指出具体修改行。
+- 环境风险：仅 Python 改动，复现成本低于含 C++ kernel 的样本。
+- flaky 风险：`atleast_nd` 为确定性张量操作，flaky 风险低；verifier 仍应重复运行抽取稳定 F2P/P2P nodeid。
+- 拆分风险：该 PR 的目标集中在新增三个 atleast API，适合作为一个样本。

swe-paddle/tasks/PaddlePaddle__Paddle-58323/solution/code.patch

@@ -0,0 +1,237 @@
+diff --git a/python/paddle/__init__.py b/python/paddle/__init__.py
+index 92ac2fcbb5..c71af71396 100644
+--- a/python/paddle/__init__.py
++++ b/python/paddle/__init__.py
+@@ -199,6 +199,9 @@ from .tensor.logic import (  # noqa: F401
+
+
+ from .tensor.manipulation import (  # noqa: F401
++    atleast_1d,
++    atleast_2d,
++    atleast_3d,
+     cast,
+     cast_,
+     concat,
+@@ -833,6 +836,9 @@ __all__ = [
+     'logspace',
+     'reshape',
+     'reshape_',
++    'atleast_1d',
++    'atleast_2d',
++    'atleast_3d',
+     'reverse',
+     'nonzero',
+     'CUDAPinnedPlace',
+diff --git a/python/paddle/tensor/__init__.py b/python/paddle/tensor/__init__.py
+index 1d6d6f8a4f..a1557bb458 100644
+--- a/python/paddle/tensor/__init__.py
++++ b/python/paddle/tensor/__init__.py
+@@ -112,6 +112,9 @@ from .logic import allclose  # noqa: F401
+ from .logic import isclose  # noqa: F401
+ from .logic import equal_all  # noqa: F401
+ from .logic import is_tensor  # noqa: F401
++from .manipulation import atleast_1d  # noqa: F401
++from .manipulation import atleast_2d  # noqa: F401
++from .manipulation import atleast_3d  # noqa: F401
+ from .manipulation import cast  # noqa: F401
+ from .manipulation import cast_  # noqa: F401
+ from .manipulation import concat  # noqa: F401
+@@ -731,6 +734,9 @@ tensor_method_func = [
+     'normal_',
+     'index_fill',
+     'index_fill_',
++    'atleast_1d',
++    'atleast_2d',
++    'atleast_3d',
+ ]
+
+ # this list used in math_op_patch.py for magic_method bind
+diff --git a/python/paddle/tensor/manipulation.py b/python/paddle/tensor/manipulation.py
+index 0686a17fa9..3195ffe8a8 100644
+--- a/python/paddle/tensor/manipulation.py
++++ b/python/paddle/tensor/manipulation.py
+@@ -3955,6 +3955,184 @@ def reshape_(x, shape, name=None):
+         return out
+
+
++def atleast_1d(*inputs, name=None):
++    """
++    Convert inputs to tensors and return the view with at least 1-dimension. Scalar inputs are converted,
++    one or high-dimensional inputs are preserved.
++
++    Args:
++        inputs (Tensor|list(Tensor)): One or more tensors. The data type is ``float16``, ``float32``, ``float64``, ``int16``, ``int32``, ``int64``, ``int8``, ``uint8``, ``complex64``, ``complex128``, ``bfloat16`` or ``bool``.
++        name (str, optional): Name for the operation (optional, default is None). For more information, please refer to :ref:`api_guide_Name`.
++
++    Returns:
++        One Tensor, if there is only one input.
++        List of Tensors, if there are more than one inputs.
++
++    Examples:
++        .. code-block:: python
++
++            >>> import paddle
++
++            >>> # one input
++            >>> x = paddle.to_tensor(123, dtype='int32')
++            >>> out = paddle.atleast_1d(x)
++            >>> print(out)
++            Tensor(shape=[1], dtype=int32, place=Place(cpu), stop_gradient=True,
++            [123])
++
++            >>> # more than one inputs
++            >>> x = paddle.to_tensor(123, dtype='int32')
++            >>> y = paddle.to_tensor([1.23], dtype='float32')
++            >>> out = paddle.atleast_1d(x, y)
++            >>> print(out)
++            [Tensor(shape=[1], dtype=int32, place=Place(cpu), stop_gradient=True,
++            [123]), Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True,
++            [1.23000002])]
++
++            >>> # more than 1-D input
++            >>> x = paddle.to_tensor(123, dtype='int32')
++            >>> y = paddle.to_tensor([[1.23]], dtype='float32')
++            >>> out = paddle.atleast_1d(x, y)
++            >>> print(out)
++            [Tensor(shape=[1], dtype=int32, place=Place(cpu), stop_gradient=True,
++            [123]), Tensor(shape=[1, 1], dtype=float32, place=Place(cpu), stop_gradient=True,
++            [[1.23000002]])]
++    """
++    out = []
++    for tensor in inputs:
++        tensor = paddle.to_tensor(tensor)
++        if tensor.dim() == 0:
++            result = tensor.reshape((1,))
++        else:
++            result = tensor
++        out.append(result)
++
++    if len(out) == 1:
++        return out[0]
++    else:
++        return out
++
++
++def atleast_2d(*inputs, name=None):
++    """
++    Convert inputs to tensors and return the view with at least 2-dimension. Two or high-dimensional inputs are preserved.
++
++    Args:
++        inputs (Tensor|list(Tensor)): One or more tensors. The data type is ``float16``, ``float32``, ``float64``, ``int16``, ``int32``, ``int64``, ``int8``, ``uint8``, ``complex64``, ``complex128``, ``bfloat16`` or ``bool``.
++        name (str, optional): Name for the operation (optional, default is None). For more information, please refer to :ref:`api_guide_Name`.
++
++    Returns:
++        One Tensor, if there is only one input.
++        List of Tensors, if there are more than one inputs.
++
++    Examples:
++        .. code-block:: python
++
++            >>> import paddle
++
++            >>> # one input
++            >>> x = paddle.to_tensor(123, dtype='int32')
++            >>> out = paddle.atleast_2d(x)
++            >>> print(out)
++            Tensor(shape=[1, 1], dtype=int32, place=Place(cpu), stop_gradient=True,
++            [[123]])
++
++            >>> # more than one inputs
++            >>> x = paddle.to_tensor(123, dtype='int32')
++            >>> y = paddle.to_tensor([1.23], dtype='float32')
++            >>> out = paddle.atleast_2d(x, y)
++            >>> print(out)
++            [Tensor(shape=[1, 1], dtype=int32, place=Place(cpu), stop_gradient=True,
++            [[123]]), Tensor(shape=[1, 1], dtype=float32, place=Place(cpu), stop_gradient=True,
++            [[1.23000002]])]
++
++            >>> # more than 2-D input
++            >>> x = paddle.to_tensor(123, dtype='int32')
++            >>> y = paddle.to_tensor([[[1.23]]], dtype='float32')
++            >>> out = paddle.atleast_2d(x, y)
++            >>> print(out)
++            [Tensor(shape=[1, 1], dtype=int32, place=Place(cpu), stop_gradient=True,
++            [[123]]), Tensor(shape=[1, 1, 1], dtype=float32, place=Place(cpu), stop_gradient=True,
++            [[[1.23000002]]])]
++    """
++    out = []
++    for tensor in inputs:
++        tensor = paddle.to_tensor(tensor)
++        if tensor.dim() == 0:
++            result = tensor.reshape((1, 1))
++        elif tensor.dim() == 1:
++            result = paddle.unsqueeze(tensor, axis=0)
++        else:
++            result = tensor
++        out.append(result)
++
++    if len(out) == 1:
++        return out[0]
++    else:
++        return out
++
++
++def atleast_3d(*inputs, name=None):
++    """
++    Convert inputs to tensors and return the view with at least 3-dimension. Three or high-dimensional inputs are preserved.
++
++    Args:
++        inputs (Tensor|list(Tensor)): One or more tensors. The data type is ``float16``, ``float32``, ``float64``, ``int16``, ``int32``, ``int64``, ``int8``, ``uint8``, ``complex64``, ``complex128``, ``bfloat16`` or ``bool``.
++        name (str, optional): Name for the operation (optional, default is None). For more information, please refer to :ref:`api_guide_Name`.
++
++    Returns:
++        One Tensor, if there is only one input.
++        List of Tensors, if there are more than one inputs.
++
++    Examples:
++        .. code-block:: python
++
++            >>> import paddle
++
++            >>> # one input
++            >>> x = paddle.to_tensor(123, dtype='int32')
++            >>> out = paddle.atleast_3d(x)
++            >>> print(out)
++            Tensor(shape=[1, 1, 1], dtype=int32, place=Place(cpu), stop_gradient=True,
++            [[[123]]])
++
++            >>> # more than one inputs
++            >>> x = paddle.to_tensor(123, dtype='int32')
++            >>> y = paddle.to_tensor([1.23], dtype='float32')
++            >>> out = paddle.atleast_3d(x, y)
++            >>> print(out)
++            [Tensor(shape=[1, 1, 1], dtype=int32, place=Place(cpu), stop_gradient=True,
++            [[[123]]]), Tensor(shape=[1, 1, 1], dtype=float32, place=Place(cpu), stop_gradient=True,
++            [[[1.23000002]]])]
++
++            >>> # more than 3-D input
++            >>> x = paddle.to_tensor(123, dtype='int32')
++            >>> y = paddle.to_tensor([[[[1.23]]]], dtype='float32')
++            >>> out = paddle.atleast_3d(x, y)
++            >>> print(out)
++            [Tensor(shape=[1, 1, 1], dtype=int32, place=Place(cpu), stop_gradient=True,
++            [[[123]]]), Tensor(shape=[1, 1, 1, 1], dtype=float32, place=Place(cpu), stop_gradient=True,
++            [[[[1.23000002]]]])]
++    """
++    out = []
++    for tensor in inputs:
++        tensor = paddle.to_tensor(tensor)
++        if tensor.dim() == 0:
++            result = tensor.reshape((1, 1, 1))
++        elif tensor.dim() == 1:
++            result = paddle.unsqueeze(tensor, axis=[0, 2])
++        elif tensor.dim() == 2:
++            result = paddle.unsqueeze(tensor, axis=2)
++        else:
++            result = tensor
++        out.append(result)
++
++    if len(out) == 1:
++        return out[0]
++    else:
++        return out
++
++
+ def gather_nd(x, index, name=None):
+     """
+