Adaptnlp

Updated to Transformers 2.7.0 which includes the Bart and T5 Language Models!

`EasySummarizer` 47
- New module for summarizing documents. These support both the T5 and Bart pre-trained models provided by Hugging Face.
- Helper objects for the easy module that can be run as standalone instances
`TransformersSummarizer`

`EasyTranslator` 49
- New module for translating documents with T5 pre-trained models provided by Hugging Face.
- Helper objects for the easy module that can be run as standalone instances
`TransformersTranslator`

Documentation and Tutorials 52
- New Class API documentation for `EasySummarizer` and `EasyTranslator`
- New tutorial guides, initial notebooks, and links to colab for the above as well
- Readme provides quickstart samples that show examples from the notebooks 53

Other
- Dockerhub repo for adaptnlp-rest added here https://hub.docker.com/r/achangnovetta/adaptnlp-rest
- Upgraded CircleCI allowing us to run 40
- Added Nightly build 39

Sequence Classification and Question Answering updates to integrate Hugging Face's public models.

`EasySequenceClassifier`
- Can now take Flair and Transformers pre-trained sequence classification models as input in the `model_name_or_path` param
- Helper objects for the easy module that can be run as standalone instances
`TransformersSequenceClassifier`
`FlairSequenceClassifier`

`EasyQuestionAnswering`
- Can now take Transformers pre-trained sequence classification models as input in the `model_name_or_path` param
- Helper objects for the easy module that can be run as standalone instances
`TransformersQuestionAnswering`

Documentation and Tutorials
Documentation has been updated with the above implementations
- Tutorials updated with better examples to convey changes
- Class API docs updated
- Tutorial notebooks updated
- Colab notebooks better displayed on readme

FastAPI Rest

AdaptNLP's first published release on github.

Easy API:
- `EasyTokenTagger`
- `EasySequenceClassifier`
- `EasyWordEmbeddings`
- `EasyStackedEmbeddings`
- `EasyDocumentEmbeddings`

Training and Fine-tuning Interface
- `SequenceClassifierTrainer`
- `LMFineTuner`

FastAPI AdaptNLP App for Streamlined Rapid NLP-Model Deployment
- adaptnlp/rest
- configured to run any pretrained and custom trained flair/adaptnlp models
- compatible with nvidia-docker for GPU use
- AdaptNLP integration but loosely coupled

Documentation
- Documentation release with walk-through guides, tutorials, and Class API docs of the above
- Built with mkdocs, material for mkdocs, and mkautodoc

Tutorials
- IPython/Colab Notebooks provided and updated to showcase AdaptNLP Modules

Continuous Integration
- CircleCI build and tests running successfully and minimally
- Github workflow for pypi publishing added

Formatting
- Flake8 and Black adherence