Release v0.7.2

Very minor, only readme/docs update.

Release v0.7.1

Documentation is now hosted on RTD.

Release 0.7.0

- Trim down `requirements.txt`
- Add a pre-commit hook to format `.py` files with [yapf](https://github.com/google/yapf)
- Add shape comments whenever possible (closes #9)
- Add `batch_first` parameter (closes #11)
- Move dimension checking stuff to a separate method (closes #8)
- Upgrade to PyTorch v1.0.0
- Add new reduction methods for `.forward()`
- Huge refactor

Release 0.6.0

* PyTorch 0.4.1 support
* Batched decoding (thanks @JeppeHallgren)

Release 0.5.1

* Add __repr__ dunder method
* Fix `forward` docstring
* Other minor fixes

Release 0.5.0

* Remove support for keyword argument `summed` in `forward`
* Remove unneeded asterisks when creating tensors
* Convert mask to long before summing to avoid overflow
* Set status to beta

Release 0.4.1

* Initialize parameters ini `__init__` (fixes #1)
* Refactor tests
* Deprecate `summed` in favor of `reduce` (fixes #2)
* Rename setup.cfg to .flake8

Release 0.3.2

* Backport from 0.4.1 (fixes #1 and #2)
* Some refactorings

Version 0.4.0

* Upgrade PyTorch to 0.3.0
* Use fancy indexing when calculating transition score
* Make squeeze dimension explicit to avoid bugs

Version 0.3.1

* Fix error when batch size is one