Zecong Hu

> That seems like the common semantics for other parameters, where do you expect things to break backwards-compatibility? I agree it's common semantics for other parameters, but the current state...

Right, that's how I would change it as well. I actually had a patch locally that looks like this — a very simple change: ```diff diff --git a/src/attr/_make.py b/src/attr/_make.py index...

No worries! Sorry for not explaining this properly. You're right that in my example, `a.kw_only` doesn't change unless being redefined. However, that is _not_ what's happening in the current implementation:...

Just wondering if there's still planned support for overriding the class `kw_only=True` setting in attributes? I don't think this is supported as of v21.2.0.

I agree that their implementations are a bit different and it's more reliable to use their code as is, but I don't think I would approve of merging into master...

1. `PairedTextData` directly reads from files. Please take a look at its [hparams](https://texar-pytorch.readthedocs.io/en/latest/code/data.html#texar.torch.data.PairedTextData.default_hparams), where you can set file paths for both the source-side and target-side datasets. 2. (and 3.) I'm...

Thank you for your feedback! These are all valuable suggestions and I think we could add them. We're actually discussing the possibility to deprecate the `Vocab` class and switch everything...

Hi, thanks for the issue! Since the example is ported from Texar-TF, I just checked the original code and indeed only `h` is used. Sorry for this careless mistake, I...

You might also want to change the PR title; should reference #166 instead.

I'm assuming by multi-GPU you mean something similar to `DataParallel` or `DistributedDataParallel`. We support multi-GPU to some extent -- The modules we provide are generally device agnostic so they could...