yusuketomoto/chainer-fast-neuralstyle

Torch Implementation of paper [Not an issue]

hashbangCoder opened this issue · 4 comments

hashbangCoder commented

Hey everyone,

I have implemented the paper in torch and have some preliminary results put up here
Please check it out if you can. Any suggestion/idea/ feedback is welcome. I'm still tuning batchsize, hyper-params and image size and currently training different models.

Sorry for creating a new issue here :/

ttoinou commented

Hi ! Thanks for the new implementation, I'll test it. Are models compatible between the two implementations ?

hashbangCoder commented

Im not sure what you mean, but if you're asking about compatibility between the torch and chainer version, then no. This is a standalone implementation in torch

mxchinegod commented

They're talking about the fact that models from the older version of yusuketomoto's repo don't work and you must use more recent models for the new code. Don't want to speak for them though and I could be wrong.

logic1988 commented

Which one was more faster? include training and generating?