[Accuracy diff No.79] Fix accuracy diff for paddle.combinations API

[ooooo-create]

PR Category

Execute Infrastructure

PR Types

Improvements

Description

日志显示反向梯度为 None，没有梯度回传
删除部分特判 paddle.empty(shape=[0, r], dtype=x.dtype) 代码，实际上去掉代码的运行结果还是正确的（torch 代码上也没有特判https://github.com/pytorch/pytorch/blob/d632cf2cc9aac8ab0e03d1537982265e42be95e5/aten/src/ATen/native/Itertools.cpp#L60-L73

• PadddleAPITest 测试通过
  

TODO

• GPU 环境下单测失败，怀疑是 masked_select_grad 的 0 size 支持问题，问题反馈到 [0-size Tensor No.132、236、240、241、259] Add 0-size Tensor support for masked_select #73180 等修复之后试试看能不能解决

[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.

[ooooo-create]

/re-run all-failed