Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

about contrib.sym.ctc_loss? #7

Closed
BackT0TheFuture opened this issue Aug 15, 2017 · 6 comments
Closed

about contrib.sym.ctc_loss? #7

BackT0TheFuture opened this issue Aug 15, 2017 · 6 comments

Comments

@BackT0TheFuture
Copy link

Hi,there
how to change warpctc with mx.contrib.sym.ctc_loss?
thx!

@Soonhwan-Kwon
Copy link
Member

Okay I expect it easy and will try to update it today or tomorrow

@Soonhwan-Kwon
Copy link
Member

Soonhwan-Kwon commented Aug 16, 2017

I tried to make change to mx.contrib.sym.ctc_loss but I found that there is an issue for performance now.
Please see apache/mxnet#7154.
As far as I know baidu's warpctc has gpu mode, so cpu mode for ctc_loss is inappropriate.
Meanwhile, I expect external warpctc is safe to use.

@BackT0TheFuture
Copy link
Author

got it !
thanks !

@fanlu
Copy link

fanlu commented Nov 3, 2017

apache/mxnet#7442
Dose this not work?

@Soonhwan-Kwon
Copy link
Member

Soonhwan-Kwon commented Nov 3, 2017

It seems that there is a problem still.
Please refer below link
apache/mxnet#7445.
And it seems cpu version and cudnn7 version can not be equal implementation,
and they plans to turn cudnn CTC off because of the API design.
@fanlu

@fanlu
Copy link

fanlu commented Nov 3, 2017

It looks like a problem.But how do you deal with problems of mx.sym.Group([softmax_loss, ctc_loss]) and STTMetric? @Soonhwan-Kwon

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants