Vladimir Iashin

I think I was able to get the diffs with another approach. Specifically, for my use case, it was enough to see the difference between two files that would be...

Hi, there was some problem with our data storage provider during the last few days. It should be resolved by now. Please try to download them again. It seems like...

Hi, Yes, it would definitely be better. Please, form a PR if you know the way.

Just propagate attention https://github.com/v-iashin/BMT/blob/d45ad8f11d35bc7d1757ac1d735952384b4fa4c1/model/multihead_attention.py#L19 to the output of the modules and apply an aggregation function of your liking.

I meant that you will need to add this variable to the output of each module.

Hi, issue starter! It is easier than you think. The features are extracted for the whole video regardless of the dataset part. During training, we can simply trim the feature...

The set of default hyper-parameters require a 12 GB GPU (1080Ti/2080Ti)

Hi. Thanks for the interest in our work! I am afraid there are several obstacles that are hard to solve before fulfilling your request, namely: 1) Google ASR system used...

No problem. Thanks for your questions. 1. Well, it doesn't matter if you use the features stored in a `.hdf5` format or as a separate file. One benefit of storing...

Great 🎉! Yep, I also noticed it today 🙂. `.size()` should fail with an error as numpy.ndarray doesn't have `size` attribute (but `pytorch` has). I don't remember if there were...