Open Asthestarsfalll opened 1 month ago
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.
@cyber-pioneer 老师大部分CI已通过,请先审一下吧。
Sorry to inform you that 7c9e1dd's CIs have passed for more than 7 days. To prevent PR conflicts, you need to re-run all CIs manually.
Sorry to inform you that d1f4508's CIs have passed for more than 7 days. To prevent PR conflicts, you need to re-run all CIs manually.
缺少float16和float64的单测?
任务完成度很高,手动点赞👍
@Asthestarsfalll 需要附上 RFC 链接。如果之前的RFC思路和你这儿的有所区别,需要提一个修改 RFC 的PR
@Asthestarsfalll 需要附上 RFC 链接。如果之前的RFC思路和你这儿的有所区别,需要提一个修改 RFC 的PR
已添加~
Sorry to inform you that 1cba1b5's CIs have passed for more than 7 days. To prevent PR conflicts, you need to re-run all CIs manually.
@cyber-pioneer 老师请审核~
PR Category
Others
PR Types
Others
Description
rfc:
中文文档:
lp pool,前向公式
$$ f(X) = (\sum_{x \in X} x^p)^{\frac{1}{p}} $$
反向公式
$$ grad = \sum_{x \in X} (x^{p - 1} f(X)^{1 - p}) $$
本实现直接借用了
PoolRawKernel
,当norm_type
为 $\infty$ 时,等价与最大池化,因此直接调用了MaxPool
;这样似乎可以直接支持LPPool3D,如果有需求的话。单测结果: