Wellcomeml

Release date: 25/02/2021

Due to an error, some of the improvements in the previous release weren't built in the PyPI whl.

- 233 CNN predict when X tf.data.Dataset
- 214 Configurable logs
- 204 Efficient CNN enabling tensorflow datasets as input
- 193 TransfomersTokenizer class

Bug fixes:
- Disables HDBScan for clustering, to solve 197.

Release date: 24/02/2021

From now on we will be using Semantic Versioning, instead of the previous CalVer system. Previous versions will still be installabe by pinning it via pip (e.g. `pip install wellcomeml==2021.2.0`), but won't be listed in the PyPI registry anymore. See discussion [here](https://github.com/wellcometrust/WellcomeML/issues/202) for more details.

Improvements
- 211 Batch prediction for `wellcomeml.ml.SemanticSimilarityClassifier`
- ~~214 Configurable logs~~ only avaliable in 1.0.2
- ~~204 Efficient CNN enabling tensorflow datasets as input~~ only available in 1.0.2
- ~~193 TransfomersTokenizer class~~ only available in 1.0.2

Bug fixes:
- ~~Disables HDBScan for clustering, to solve 197.~~

Notice: for a technical reason, we are starting with 1.0.1 and not 1.0.0