working_links.txt

https://github.com/cxa-unique/Simplified-TinyBERT
https://github.com/yinmingjun/TinyBERT/tree/e4f2e27a24431f1b90b2d94c8a10e847ababab6e
https://hatespeechdata.com/#English-header - hatespeech datasets

commands
python pregenerate_training_data.py --train_corpus ..\data\enwiki_0.txt --bert_model bert-base-uncased --reduce_memory --do_lower_case --epochs_to_generate 3 --output_dir ..\output_data 


Next steps:
1) take general distilled TinyBert and try to task distill on hate speech or so - create pipeline
------> Create a abstraction for BertForSequenceClassification in modeling.py and then create BertFineTuning
        Script.
------> Test the running of the pipeline.

2) Then try to create analysis pipeline where we analyze the attentions