This is my solution for the Kaggle Real or Not? NLP with disaster tweets getting started competition.
pip install -r requirements.txt
All commands are executed using
python main.py
and using the flags
Preprocessing carries out some cleanup of the data. Run using:
python main.py --mode=preprocess
First run the BERT preprocessing
python main.py --mode=preprocess_bert
This will generate the training data in a python shelve file created at ./tmp/bert_data
.
Next run the training program
python main.py --mode=train_bert