Kyle Gorman

icon indicating a website link wellformedness.com icon indicating an email [email protected]

icon company Graduate Center, City University of New York; @Google AI icon location Brooklyn, NY icon location Computational linguist, professor; phonology and morphology, language resources, finite-state technologies, neural networks.

Results 183 comments of Kyle Gorman

Consistent batch size

Now that #40 is closed and such, I am wondering: what is the best way to enable this? * I think it should be a flag `--pad_max` which causes the...

Consistent batch size

Okay, so: * `--pad_max` is the flag. * `Dataset` computes the max(es) * This gets passed as the `pad_max` argument in `batches.PaddedTensor` calls via the collators. * I may add...

Consistent batch size

I now have an early draft of this [here](https://github.com/kylebgorman/yoyodyne/tree/max/yoyodyne), in the `max` branch of my fork. Important simplifications while I get the basics right: * I assume that we are...