PaddlePaddle · Echo-Nie · Jul 22, 2026
diff --git a/swe-paddle/tasks/PaddlePaddle__Paddle-79161/README.md b/swe-paddle/tasks/PaddlePaddle__Paddle-79161/README.md
@@ -0,0 +1,44 @@
+# PaddlePaddle__Paddle-79161
+
+This directory converts Paddle PR #79161 into a SWE-Paddle community task candidate.
+
+## Source
+
+| Field | Value |
+| --- | --- |
+| Repo | `PaddlePaddle/Paddle` |
+| PR | [79161](https://github.com/PaddlePaddle/Paddle/pull/79161) |
+| PR title | `[API Compatibility] Add param alias for paddle.set_rng_state` |
+| Base commit | `8dd02b271734f7aae3669fe6dbcbea57d9cc9add` |
+| Merged at | `2026-05-28` |
+| Task type | `feature_enhancement` |
+| Resource | CPU |
+
+## Summary
+
+Add a compatible alias `new_state` for the existing `state_list` parameter of `paddle.set_rng_state`, while keeping the original calling method unchanged.
+
+## Why This Is A Good SWE-Paddle Candidate
+
+- It adds a user-visible API compatibility capability rather than repairing an internal-only failure.
+- The production change is Python-only and limited to one public random-state API.
+- Independent behavior tests can distinguish Base and Solution on CPU without source build, GPU, or external data.
+- Positional and original keyword calls provide clear P2P coverage, while alias dispatch and conflict handling provide direct F2P signals.
+
+## 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.
+- `README.md`: task overview and verification entrypoint.
+
+## 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.
diff --git a/swe-paddle/tasks/PaddlePaddle__Paddle-79161/environment/README.md b/swe-paddle/tasks/PaddlePaddle__Paddle-79161/environment/README.md
@@ -0,0 +1,27 @@
+# Environment Notes
+
+This candidate is part of the SWE-Paddle community task set.
+
+## Expected Environment
+
+- Repository: `PaddlePaddle/Paddle`
+- Base commit: `8dd02b271734f7aae3669fe6dbcbea57d9cc9add`
+- Resource: CPU
+- GPU required: no
+- Build path: Python-only checkout source with an AST overlay for the target random-state function; no Paddle source build is required.
+
+## 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.
diff --git a/swe-paddle/tasks/PaddlePaddle__Paddle-79161/instruction.md b/swe-paddle/tasks/PaddlePaddle__Paddle-79161/instruction.md
@@ -0,0 +1,17 @@
+# 为 paddle.set_rng_state 增加 new_state 参数别名
+
+## 详细描述
+
+`paddle.set_rng_state` 当前仅接受 `state_list` 作为随机数生成器状态参数。使用其他框架风格代码或自动迁移工具时，调用方可能使用语义等价的 `new_state` 参数，导致现有 API 无法直接兼容。该 API 应支持 `new_state` 作为 `state_list` 的等价别名，同时保持原有调用方式和错误语义稳定。
+
+## 验收说明
+
+- `paddle.set_rng_state(new_state=...)` 应与使用 `state_list=...` 产生相同的状态设置行为。
+- positional 参数和原有 `state_list=` 关键字调用应保持兼容。
+- 同时传入 `state_list` 与 `new_state` 时，应明确拒绝冲突输入。
+
+## 技术要求
+
+- 熟悉 Python function signature 和 decorator 行为。
+- 了解 Paddle random generator state API。
+- 能够编写稳定的 API compatibility behavior tests。
diff --git a/swe-paddle/tasks/PaddlePaddle__Paddle-79161/proposal.md b/swe-paddle/tasks/PaddlePaddle__Paddle-79161/proposal.md
@@ -0,0 +1,55 @@
+# Task Proposal: PaddlePaddle__Paddle-79161
+
+## 1. 来源信息
+
+- Instance ID：`PaddlePaddle__Paddle-79161`
+- PR 链接：https://github.com/PaddlePaddle/Paddle/pull/79161
+- PR 标题：`[API Compatibility] Add param alias for paddle.set_rng_state`
+- `base_commit`：`8dd02b271734f7aae3669fe6dbcbea57d9cc9add`
+- merged 时间：`2026-05-28`
+- 你的身份：熟悉该模块的 contributor
+- 后续联系人：TBD
+
+## 2. 问题一句话
+
+为 `paddle.set_rng_state` 的 `state_list` 参数增加等价别名 `new_state`，提升与其他框架风格调用的兼容性。
+
+## 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 API Compatibility PR，不是合成任务。
+- **代表性**：它覆盖公共 Python API 的参数 alias、向后兼容和冲突参数语义。
+- **边界清楚**：production 行为集中在 `paddle.set_rng_state` 的调用签名兼容性。
+- **非平凡性**：实现不仅要接受新 alias，还必须保留 positional 与原关键字调用，并正确拒绝同时传入 canonical 参数和 alias 的冲突场景。
+- **环境友好性**：目标逻辑为 Python-only，可通过 checkout source AST overlay 在 CPU 环境稳定验证，无需 source build。
+
+## 4. 任务类型和标签
+
+- 任务类型：`feature_enhancement`
+- 执行后端：`cpu`
+- 设备范围：`cpu_only`
+- 模块标签：`[random, api_compatibility, parameter_alias, decorator, python_only]`
+
+## 5. 验证思路
+
+- 目标测试命令：`bash tests/test.sh`
+- 目标测试文件：`test/swe_paddle/test_pr79161_set_rng_state_alias.py`
+- 修复前预期：positional 和 `state_list=` 既有调用通过；`new_state=` alias 与 alias 冲突语义测试失败。
+- 修复后预期：继续应用 `solution/code.patch` 后，全部目标测试通过。
+- P2P 候选：positional state list、原 `state_list=` 关键字和显式 CPU device 参数的既有行为。
+
+## 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
+- patch 类型：Python-only
+- 环境建议：从 checkout source 提取 `set_rng_state` 的真实 function 与 decorator control flow，并以 controlled CPU place 和 generator doubles 观察 state side effect，无需导入完整历史 Paddle 模块。
+- 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：`instruction.md` 仅描述公共 API 行为，不指出具体 decorator 或修改行。
+- 环境风险：AST overlay 避免历史 source 与当前 Paddle wheel 的整体 import 兼容问题。
+- flaky 风险：测试只验证确定性的参数映射、异常类型和 CPU generator side effect，不依赖随机采样结果。
+- 拆分风险：alias 支持及其冲突语义属于同一个公共 API compatibility 目标，适合作为一个样本。
diff --git a/swe-paddle/tasks/PaddlePaddle__Paddle-79161/solution/code.patch b/swe-paddle/tasks/PaddlePaddle__Paddle-79161/solution/code.patch
@@ -0,0 +1,28 @@
+diff --git a/python/paddle/framework/random.py b/python/paddle/framework/random.py
+index d623182e024674e2da74690307417d757d465268..f00353b5d973ae3449e85420740932a375fb3a0c 100644
+--- a/python/paddle/framework/random.py
++++ b/python/paddle/framework/random.py
+@@ -18,6 +18,7 @@ from typing import TYPE_CHECKING
+
+ import paddle
+ from paddle.base import core
++from paddle.utils.decorator_utils import param_one_alias
+
+ if TYPE_CHECKING:
+     from collections.abc import Sequence
+@@ -150,6 +151,7 @@ def get_cuda_rng_state() -> list[paddle.base.core.GeneratorState]:
+     return state_list
+
+
++@param_one_alias(["state_list", "new_state"])
+ def set_rng_state(
+     state_list: Sequence[paddle.base.core.GeneratorState],
+     device: str | None = None,
+@@ -160,6 +162,7 @@ def set_rng_state(
+
+     Args:
+         state_list(list|tuple): The device states to set back to device generators. state_list is obtained from get_rng_state().
++            Alias: ``new_state``.
+         device(str): This parameter determines the specific running device.
+             It can be ``cpu``, ``gpu``, ``xpu``, Default is None.
+             If None, return the generators of current device (specified by ``set_device``).
diff --git a/swe-paddle/tasks/PaddlePaddle__Paddle-79161/tests/test.patch b/swe-paddle/tasks/PaddlePaddle__Paddle-79161/tests/test.patch
@@ -0,0 +1,146 @@
+diff --git a/test/swe_paddle/test_pr79161_set_rng_state_alias.py b/test/swe_paddle/test_pr79161_set_rng_state_alias.py
+new file mode 100644
+index 0000000..11ceea9
+--- /dev/null
++++ b/test/swe_paddle/test_pr79161_set_rng_state_alias.py
+@@ -0,0 +1,140 @@
++import ast
++import copy
++import functools
++from pathlib import Path
++from types import SimpleNamespace
++
++import pytest
++
++
++TARGET = Path("python/paddle/framework/random.py")
++
++
++class _CPUPlace:
++    pass
++
++
++class _CUDAPlace:
++    pass
++
++
++class _XPUPlace:
++    pass
++
++
++class _CustomPlace:
++    pass
++
++
++class _Generator:
++    def __init__(self):
++        self.states = []
++
++    def set_state(self, state):
++        self.states.append(state)
++
++
++def _param_one_alias(names):
++    canonical, alias = names
++
++    def decorator(function):
++        @functools.wraps(function)
++        def wrapper(*args, **kwargs):
++            if alias in kwargs:
++                if canonical in kwargs:
++                    raise ValueError(
++                        f"Cannot specify both '{canonical}' and its alias '{alias}'."
++                    )
++                kwargs[canonical] = kwargs.pop(alias)
++            return function(*args, **kwargs)
++
++        return wrapper
++
++    return decorator
++
++
++class _StripAnnotations(ast.NodeTransformer):
++    def visit_arg(self, node):
++        node.annotation = None
++        return node
++
++    def visit_FunctionDef(self, node):
++        self.generic_visit(node)
++        node.returns = None
++        return node
++
++
++def _load_checkout_function():
++    tree = ast.parse(TARGET.read_text(encoding="utf-8"), filename=str(TARGET))
++    function = next(
++        node
++        for node in tree.body
++        if isinstance(node, ast.FunctionDef) and node.name == "set_rng_state"
++    )
++    function = _StripAnnotations().visit(copy.deepcopy(function))
++    ast.fix_missing_locations(function)
++
++    generator = _Generator()
++    converted_devices = []
++
++    def convert_to_place(device):
++        converted_devices.append(device)
++        return _CPUPlace()
++
++    core = SimpleNamespace(
++        CPUPlace=_CPUPlace,
++        default_cpu_generator=lambda: generator,
++        get_cuda_device_count=lambda: 0,
++        get_xpu_device_count=lambda: 0,
++        get_all_custom_device_type=lambda: [],
++        get_custom_device_count=lambda _device_type: 0,
++    )
++    paddle = SimpleNamespace(
++        CPUPlace=_CPUPlace,
++        CUDAPlace=_CUDAPlace,
++        XPUPlace=_XPUPlace,
++        CustomPlace=_CustomPlace,
++        framework=SimpleNamespace(
++            _current_expected_place_=lambda: _CPUPlace()
++        ),
++        device=SimpleNamespace(_convert_to_place=convert_to_place),
++    )
++    namespace = {
++        "core": core,
++        "paddle": paddle,
++        "param_one_alias": _param_one_alias,
++    }
++    module = ast.Module(body=[function], type_ignores=[])
++    exec(compile(module, str(TARGET), "exec"), namespace)
++    return namespace["set_rng_state"], generator, converted_devices
++
++
++def test_existing_state_list_calls_remain_valid():
++    function, generator, converted_devices = _load_checkout_function()
++    positional_state = object()
++    keyword_state = object()
++
++    function([positional_state])
++    function(state_list=[keyword_state], device="cpu")
++
++    assert generator.states == [positional_state, keyword_state]
++    assert converted_devices == ["cpu"]
++
++
++def test_new_state_alias_sets_the_same_generator_state():
++    function, generator, converted_devices = _load_checkout_function()
++    aliased_state = object()
++
++    function(new_state=[aliased_state], device="cpu")
++
++    assert generator.states == [aliased_state]
++    assert converted_devices == ["cpu"]
++
++
++def test_state_list_and_new_state_conflict_is_rejected():
++    function, generator, _ = _load_checkout_function()
++
++    with pytest.raises(ValueError, match="Cannot specify both"):
++        function(state_list=[object()], new_state=[object()])
++
++    assert generator.states == []
diff --git a/swe-paddle/tasks/PaddlePaddle__Paddle-79161/tests/test.sh b/swe-paddle/tasks/PaddlePaddle__Paddle-79161/tests/test.sh
@@ -0,0 +1,4 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+python -m pytest test/swe_paddle/test_pr79161_set_rng_state_alias.py -q