Tobias Pütz

__new__ and __init__ get no call signature

1

I am trying to add my doc-string and an explicit signature for `__new__` and `__init__` to `Foo`: ~~~rust /// Foo(arg1, arg2) /// -- /// /// A Foo #[pyclass(name=Foo)] pub struct...

torch.unique crashes on GPU

4

## 🐛 Bug Running `torch.unique` with any specified argument fails: ```py In [5]: torch.unique(torch.arange(10).view(2,5).cuda(), dim=1) [1] 17003 abort (core dumped) ipython ``` ```py In [2]: torch.unique(torch.arange(10).view(2,5).cuda(), dim=1, return_counts=True) ... RuntimeError:...

Writing the labels to yml or toml can still be done via serde but allows for manual inspection.

We could bucket sequences in the dev set by their length to speed up dev passes

Make the exponential decay lr schedule available

6

Right now, the exponential decay lr schedule is not available for `sticker train`and `sticker pretrain`. Once #145 is merged, it would make sense to have it available for both subcommands....

Add gradient norm to train output

4

Printing the gradient norm alongside the training output can be helpful when tuning learning rates and for finding the gradient clipping threshold.