Pete Walsh

Hi @borosilicate, yes please go ahead with a PR when you have a chance

@JohnGiorgi @annajung I like this idea of adding a composable decoder to CopyNet, which could also support multi-layer LSTMs. If you either of you would like to look into this,...

@JohnGiorgi one thing to keep in mind is that we want to keep backwards compatibility. So a CopyNet model trained before with a `_decoder_cell` should still work. And if the...

@brendan-ai2 will do!

Another potential solution would be to use PyTorch's tensorboard interface, instead of tensorboardX. See https://github.com/pytorch/pytorch/releases/tag/v1.1.0. Or maybe with the upcoming trainer callbacks refactoring we could also provide a callback for...

@lanpa thanks for the update! I will check it out

Related to this, is there any reason why a docstring linter such as pydocstyle shouldn't be added to the tests? I don't think pydocstyle would catch these types of errors,...

Yeup, this seems reasonable. I think we should try to automatically detect the compression type, but also have the flag so that users can override it when the automatic detection...

Hi @Dbhasin1, check out https://github.com/allenai/allennlp/blob/main/CONTRIBUTING.md#making-a-pull-request

Hi @spranjal25, we haven't heard from @Dbhasin1 for a while on their PR, so it's probably okay for you take over at this point.