Hi, thanks for the great code!

I wonder do you have plans to support resuming from checkpoints for classification? As we all know, in terms of training ImageNet, the training process is really long and it can be interrupted somehow, but I haven't notice any code related to "resume" in scripts/classification/train_imagenet.py.

Maybe @hetong007 ? Thanks in advance.

How do i resume training for text classification?

Well, Gumbel Distribution is magical. Basically, given a sequence of K logits, i.e., "\log a_1, \log a_2, ..., \log a_K" and K independent gumbel random variables, i.e., "g_1, g_2, ..., g_K". We have

\argmax_i \log a_i + g_i ~ Categorical({a_i / sum(a)})

This gives you a very simple way to sampl

We have a variety of tests scattered across the repo. @apollo-elements/mixins does most of the heavy lifting. I would like to see the tests in mixins replicated across packages. i.e. Just like ApolloQueryMixin is tested to make sure that data is set when a good query resolves, so should lit-apollo's ApolloQuery, and Polymer ApolloQuery, etc.

It would be swell to export some `testBas