Skip to content

Conversation

guoshengCS
Copy link
Contributor

resolves #3870

@qingqing01
Copy link
Contributor

qingqing01 commented Sep 5, 2017

row_l2_norm_op 想了想,觉得还是实现成一个通用的norm,在接一个elmentwise div op好一些。 TF和PyTorch的norm op如下:

这两个接口类似:

tf.norm

norm(
    tensor,
    ord='euclidean',
    axis=None,
    keep_dims=False,
    name=None
)
torch.norm(input, p, dim, keepdim=False, out=None) 

@qingqing01
Copy link
Contributor

@guoshengCS 不过这个Op既然写了,还是照常review + merge吧,norm op我们再另外写吧。 @lcy-seso @... 你们觉得的呢?

@lcy-seso
Copy link
Contributor

lcy-seso commented Sep 5, 2017

我的意见是花一些时间改成 P 范数, p = 1,2。是否需要考虑更多情况,@guoshengCS 可以再自行考虑一下~

@guoshengCS guoshengCS mentioned this pull request Sep 29, 2017
@qingqing01 qingqing01 closed this Oct 23, 2017
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

RowL2Norm Operator

3 participants