Greg Bildson

I've noticed that all new instance norm models don't work when converted to coreml. Also, all historical models in fast-neural-style don't work in torch. Have yet to figure out the...

I found an older machine with an older torch setup (~ 6 months) and the server images for old models generate fine. I then stepped through the updates that I...

Yeah, batch works but lower quality and/or bigger model.

Given that I didn't mention getting better results, I don't believe that I did achieve what is mentioned in the papers.    I had a hard time getting good results in...

I have the qualify.bin as output from this: https://github.com/gbildson/nprizeadditions/blob/master/other/qualify2bin.py but I don't have the qualify.txt handy anywhere. It appears to be a binary representation of qualify.txt. Would that help? Netflix...

Why would you "modularize" components but then not support their ops in tensorflow-serving?

Hmmm, possibly. Does that appear to work better? I think I found random init working better. I've realized a while ago that my softmax calc was off given better understanding...