ELECTRA model support added for NER tasks

Added

• Added support for ELECTRA based NER models.

Added support for proxies with ConvAI

Fixed

• Fixed bug in LanguageModelingModel initialization with a trained tokenizer.

Added

• Added support for passing proxy information with ConvAI model.

Bug Fixes

Fixed

• Fixed potential bug in NERModel predict() method when using custom labels.
• Fixed typo in the NERModel description in the readme.

Bug Fixes

Fixed

• Fixed issues with vocab_size not being set properly in ELECTRA models.

Bug Fixes

Fixed

• Fixed bugs in minimal examples for language modeling.

Changed

• Added vocab_size back to default args dict for clarity. (vocab_size is None by default)
• Changed error message when training a new tokenizer with incorrect parameters for clarity.

ELECTRA Pre-Training Support Added

Added

• Added ELECTRA pretraining support.
• Added better support for configuring model architectures when training language models from scratch.
  • Any options which should be overriden from the default config can now be specified in the args dict. (config key)

Changed

• Default entry for vocab_size removed from args for LanguageModelingModel as it differs for different model types.
  • vocab_size must now be specified whenever a new tokenizer is to be trained.

Fixed

• Fixed bugs when training BERT (with word piece tokenization) language models from scratch.
• Fixed incorrect special tokens being used with BERT models when training a new tokenizer.
• Fixed potential bugs with BERT tokenizer training.