Palladium

===================

- Updated requirements in order to use newer versions of dependencies
(also fixing potential security vulnerabilities in dependencies)

- Added support for handling model attachments

- Exclamation mark `!` can now be used instead of `__factory__` in
configuration files

=====================

- Added AWS S3 persister

===================

- Added default Palladium config file locations (to be checked if
`PALLADIUM_CONFIG` is not set)

- Introduced `palladium.dataset.CSV` to replace `palladium.dataset.Table`

- `__copy__` configuration directive now accepts a `__default__`

- Updated requirements in order to use newer versions of dependencies
(also fixing potential security vulnerabilities in dependencies)

=====================

- Added pld-export command to export a model from one model persister
to another.

- Added a DatasetLoader for loading data from OpenML via scikit-learn.

- Added another R example that solves a regression problem and makes
use of factor variables or categorical variables.

- Added Python 3.7 to Travis CI test settings, removed Python 3.5.

===================

- You can now run grid searches across a cluster of computers. See
the FAQ entry "Can I use my cluster to run a hyperparameter search?"
for an example.

- PredictService now has better support for text predictions (which
use 1d arrays of strings instead of 2d matrices as inputs) with the
`unwrap_sample` option.

===================

- Update use of rpy2, no longer used deprecated pandas.rpy module.

- Save `model.cv_results_` in model's metadata if it exists.

- The best model of a grid search can now be persisted using the
`--persist-best` flag.

- Grid search results can now be saved using the `--save-results`
option.

- A `/activate` endpoint allows for activating a certain model version
through the web.

- A `/refit` API endpoint that allows fitting a model to be triggered
through the web.

- A new `/list` API endpoint that works similar to the pld-list CLI

- A `scoring` variable in the configuration root, if defined, will be
used for both pld-test and pld-grid-search

- In CachedUpdatePersister don't load model if we already have active

- Allow for use of multiple CachedUpdatePersisters

Instances of CachedUpdatePersister now use their own name inside the
configuration file to index into `process_store`. That is, a
CachedUpdatePersister that's configured as `my_model_persister`
inside the configuration file will have an entry in `process_store`
with the key `my_model_persister`. This is a breaking change, since
the previously used key `model` is no longer available.

- Added a model persister for REST backends, e.g. for use with
Artifactory.

- Added a `__copy__` directive to copy and override existing entries
in configuration files. This can be used to avoid redundancy in one
or multiple configuration files.