- we already have `util` so it is confusing which util it should look for during the build

* SPARKNLP-643 Extract only the spark version(numbers)

* SPARKNLP-643 Replace regex with LooseVersion from distutils

* SPARKNLP-558 Adding serializable support to BytesKey object

* SPARKNLP-558 Adding serializable support to BytesKey object

- it should be `spm.model`

* SPARKNLP-632 Adding control for annotators with only one expected column in Python side

* SPARKNLP-632 Adding inputAnnotatorTypes and optionalInputAnnotatorTypes in annotators for Python

* SPARKNLP-632 Removing useless prints to terminal

* SPARKNLP-632 Adding AnnotatorType initialization in Python

* SPARKNLP-632 Removing wrong imports

…me models (#13048)

* code for fixing Nans

* Remove var and leave a TODO

Co-authored-by: Maziyar Panahi <maziyar.panahi@iscpif.fr>

* SPARKNLP-645 Update unit tests for CoNLLGenerator

* CoNLLGenerator Bugfixes and Improvements both on Scala and Python (#13051)

* Test for non-int metadata values in CoNLLGenerator

* Fix for non-int metadata values bug in CoNLLGenerator

* Include escaping when writing csv in order to preserve special char tokens

* Remove unnecessary option from csv write

* Addin metadata sentence key parameter in order to selecti which metadata field to use as sentence for CoNLL Generation

* Minor formatting in scala, Python refactorization (several bug fixes supporting scala overloaded methods), Python Tests for 2 and 3 arguments

Co-authored-by: Andres Fernandez <51669244+jfernandrezj@users.noreply.github.com>

#13060)

- new parameter rules where users can provide an array of rules
- new parameter delimiter to split the rules