paddle.masked_fill 在整型输入场景出现 int32/int64 类型读取不一致报错

[tingPetty]

bug描述 Describe the Bug

在最小复现中，paddle.masked_fill 报错：
ValueError: (InvalidArgument) The type of data we are trying to retrieve (int64) does not match the type of data (int32) currently contained in the container.
同样的测试在对照框架可运行成功，脚本复现标记为 REPRODUCED_PADDLE_ERROR_ONLY=True。
附上对应脚本：
pd_error_only_candidate_166_masked_fill.py

import numpy as np
import paddle

print("paddle:", paddle.__version__)

x = paddle.to_tensor(np.array([[1, 2], [3, 4]], dtype=np.int32))
mask = paddle.to_tensor(np.array([[True, False], [False, True]], dtype=np.bool_))

# 触发点
out = paddle.masked_fill(x, mask, 0)
print(out)

ValueError: (InvalidArgument) The type of data we are trying to retrieve (int64) does not match the type of data (int32) currently contained in the container.
  [Hint: Expected dtype() == phi::CppTypeToDataType<T>::Type(), but received dtype():7 != phi::CppTypeToDataType<T>::Type():9.] (at ..\paddle\phi\core\dense_tensor.cc:174)

其他补充信息 Additional Supplementary Information

• 复现环境：Windows 11，Python 3.10.18，Paddle 3.2.0，CPU。

[xuanyuanminzheng]

感谢你的问题反馈，该问题属于已知问题，低优排期修复。