QA_project - Attentive History Selection for Conversational Question Answering

Github for original TF code: https://github.com/prdwb/attentive_history_selection
Paper we are implementing in PyTorch: https://arxiv.org/abs/1908.09456

Rewritten scripts:

1. pt_cqa_supports.py python pt_cqa_supports.py
2. pt_cqa_gen_batches.py
3. bert_model.py
4. cqa_model.py - bert_rep (updated) [READY]
   - bert_segment_rep [No need to re-implement, this function is not used]
   - cqa_model [READY]
   - aux_cqa_model [No need to re-implement, this function is exactly the same as cqa_model]
   - yesno_model [READY]
   - followup_model [No need to re-implement, this function is exactly the same as yesno_model]
   - history_attention_net [READY]
   - disable_history_attention_net [READY]
   - fine_grained_history_attention_net [READY]
   

To run main script from within the 'script' directory: (15.05.2020)
python main_qa_script.py --quac_data_dir ../data/ --cache_dir ../data/final_cache_dir --output_dir ../exps_dir --batch_size 24 --do_train --fine_grained_attention