Joe Cummings

@EvilFreelancer Can you update the README with examples on how this will now look in the YAML config?

> > Unfold all params from the DictConfig (assigning them to class variables) > Not a huge fan of this. It's unnecessary because not all params will actually need to...

> I actually prefer comments in the configs - makes them more readable. And of course more comments in the recipes. Comments in the configs are a good starting point,...

> Then there's no default values? All parameters HAVE to be defined in the config. Let's say we add a new parameter controlling a new feature in a recipe. And...

> The change looks good with just some small comments around testing. Do the tests verify this works with bs > 1 too? Added this test!

> As discussed offline, my primary question here would be around the need to integrate non-incremental decoding to generation utils as opposed to just having this as a custom test...

@msaroufim This is awesome!!! Is there any way you could convert to a PR so we can comment on it line-by-line? I know there's a Google Doc, but that's only...

Hi @lordsoffallen - unfortunately, TorchText is no longer actively maintained by PyTorch. For NLP model implementations, data transforms, and datasets, I recommend HuggingFace's [Transformers](https://huggingface.co/docs/transformers/index), [Tokenizers](https://huggingface.co/docs/tokenizers/index), and [Datasets](https://huggingface.co/docs/datasets/index) libraries.

Do the following PRs need to be cherry-picked into the release? @atalman @osalpekar @seemethere - #2071 - #2075 - #2078

Okay, the above cherry-picks are ready for review and merge. Feel free to handle the ones related to RE @osalpekar @atalman