Pengzhi Gao

`GPT2Decoder` currently inherits `PretrainedGPT2Mixin`. This part might be improved (inherits `TFDecoder`) in the future.

1. Add CI test for building documentations (Do not ignore `warnings` and add spellcheck). 2. Fix docstrings with incorrect/inconsistent Sphinx format. Currently, such issues are treated as `warnings` in the...

setting `pretrained_model_name` will not only define the model arch but also load the pre-trained checkpoint. We should have another `hparam` to control whether to load pre-trained checkpoint or not.

Support ELMo in texar-pytorch

2

Requested by Forte Team. @hunterhector

Add sequence_tagging example

1

Adapted from [sequence_tagging](https://github.com/asyml/texar/tree/master/examples/sequence_tagging) in `texar-tf`.

Add ELMo modules

2

Add texar-styled ELMo encoder adapted from allennlp. The corresponding tokenizer will be in another PR. Resolve some comments in #298 I checked the implementation of `ELMo` in `allennlp`, It seems...

There are some subtle differences between `BPE` implementation in [sentencepiece](https://github.com/google/sentencepiece) and `BPE` implementation in [subword-nmt](https://github.com/rsennrich/subword-nmt). We could probably delete everthing except `multi-bleu.perl` in [texar-pytorch/bin/utils](https://github.com/asyml/texar-pytorch/tree/master/bin/utils) after this one is implemented. [Transformer...

language_model_ptb example

1

resolve #232 (including `distributed_gpu` and `language_model_ptb` examples)

https://openreview.net/pdf?id=H1eA7AEtvS

In `texar-pytorch/bin/utils`, `sentencepiece` is a package instead of a tokenization method, so `sentencepiece` encoding is not a very accurate way to describe the tokenization method. `sentencepiece` includes two sub-word tokenization...