Malaya

1. Added Coreference Resolution module, https://malaya.readthedocs.io/en/latest/load-coreference-resolution.html
2. Improved dependency parsing module, https://malaya.readthedocs.io/en/latest/load-dependency.html
<img width="1388" alt="Screenshot 2021-06-28 at 1 25 19 PM" src="https://user-images.githubusercontent.com/19810909/123591567-bb6c7d80-d81e-11eb-878d-532110c0c863.png">

1. Supported XLA for malaya models, https://malaya.readthedocs.io/en/latest/devices.html
2. Supported precision modes, can choose to run FP16, FP32 or FP64, https://malaya.readthedocs.io/en/latest/precision-mode.html
3. Improved GPU interface, https://malaya.readthedocs.io/en/latest/gpu-environment.html
4. Added Knowledge Graph Triplet transformer models, https://malaya.readthedocs.io/en/latest/load-knowledge-graph-triplet.html
5. Added Knowledge Graph parser from Dependency Parsing models, https://malaya.readthedocs.io/en/latest/load-knowledge-graph-from-dependency.html

1. Deep stemmer now support Tensorflow 2.X
2. Added distilled transformer for Abstractive Summarization, https://malaya.readthedocs.io/en/latest/load-abstractive.html#List-available-Transformer-models
3. Added Similarity module documentation.
4. Added KeyPhrase similarity Transformer model, https://malaya.readthedocs.io/en/latest/load-keyphrase-similarity.html
5. Added QA SQUAD style Transformer model, https://malaya.readthedocs.io/en/latest/load-keyphrase-similarity.html
6. Update WordVector interface, https://malaya.readthedocs.io/en/latest/load-wordvector.html#

1. Support Tensorflow 2.0! Deep stemmer model not able to use TF 2.0.

1. Added OntoNotes5 Entity recognition, https://malaya.readthedocs.io/en/latest/load-entities.html#Load-Transformer-Ontonotes-5-model
2. Added BigBird Translation, able to infer up to 1024 words, https://malaya.readthedocs.io/en/latest/load-long-text-translation.html
3. Added BigBird + Pegasus Abstractive Summarization, able to infer up to 2048 words, https://malaya.readthedocs.io/en/latest/long-text-abstractive-summarization.html
4. T5 function merged with `transformer` function.
5. Removed Bert-Tensorflow and ALBert-Tensorflow dependencies, preparing for TF2.
6. Removed contrib layers dependencies, preparing for TF2.

1. Released Kesalahan Tatabahasa module, https://malaya.readthedocs.io/en/latest/load-tatabahasa.html
2. Standardized methods for seq2seq model,
- `greedy_decoder` for greedy decoder.
- `beam_decoder` for beam decoder.
- `nucleus_decoder` for nucleus decoder.