Neural-diffeqs

What's Changed
* V0.2.0 by mvinyard in https://github.com/mvinyard/neural-diffeqs/pull/2
* V0.2.0 by mvinyard in https://github.com/mvinyard/neural-diffeqs/pull/3


**Full Changelog**: https://github.com/mvinyard/neural-diffeqs/compare/v0.2.0rc...v0.2.0rc1

- Biggest change being organization, moving the function `_compose_nn_sequential` to its own package, [`torch-composer`](https://github.com/mvinyard/torch-composer). `torch-composer` is available via:


pip install torch-composer


**Full Changelog**: https://github.com/mvinyard/neural-diffeqs/compare/v0.0.31...v0.1.0

- Updated some terminology
- Improved documentation

**Full Changelog**: https://github.com/mvinyard/neural-diffeqs/compare/v0.0.3...v0.0.31

- Added key dependencies (numpy) for sufficient self-containment

**Full Changelog**: https://github.com/mvinyard/neural-diffeqs/compare/v0.0.01...v0.0.11

* Added in-line functionalities, `psi_mu` and `psi_sigma` to the **`NeuralDiffEq`** model class such that one can call the potential directly given an input vector, without mapping it back to the input dimensionality.

**Full Changelog**: https://github.com/mvinyard/neural-diffeqs/compare/v0.0.2...v0.0.3

- Refactoring to include potential nets
- Improved documentation
- Simplified and decreased dependencies
- Bug fixes

**Full Changelog**: https://github.com/mvinyard/neural-diffeqs/compare/v0.0.11...v0.0.2