Rowan Zellers

Are you sure cuda 9.0 is installed? Looks like it might be a cuda/pytorch incompatibility issue: https://github.com/open-mmlab/mmdetection/issues/385 On Wed, Apr 24, 2019 at 1:28 AM hmyhehe wrote: > Hi, I...

oh no! I'm not sure how to solve that either, since I haven't ran into that problem before. let me know if you can find a fix :)

hey! sorry this dropped off my radar. I have some visualization code in https://github.com/rowanz/neural-motifs/blob/master/models/_visualize.py - you might also want to check the source code to my online demo at [http://rowanzellers.com/scenegraph2/](http://rowanzellers.com/scenegraph2/)...

It should be possible. The big differences might be removing the "bfloat16" casting (as that's a datatype specific to TPUs), and changing some of the boilerplate code to not be...

Hey all, I've created a script that appears to work for preprocessing. The source is [here.](https://github.com/rowanz/arctic-captions) It does everything besides create the word-ID dictionary.

Sure -- unfortunately, they're on AI2 servers I believe, and I don't work there any more so I don't have access to them -- @GloriaXimingLu do you have them somewhere?

Hi Ying, If you want the images, I think you might have to do a matching -- but hopefully there's info in there that should help, including the fields "fn"...

oh interesting! we didn't run into that though -- I think that means that all TVQA clips are longer than `segment_size`.

hi Syeda - that shouldn't be possible! I think you probably have a bug somewhere in your setup. I do have logs somewhere on wandb, but I don't think they...