Custom raceBERT

Preparation

1. make sure the folder structure remains the same

   a. data in the data folder
   b. .py files in the models folder
   c. trained_models exist to store the trained models

root  
|-- data/  
|-- models/  
|-- trainied_models/  

2. make sure the data in the data folder follows the same structure as nmzpAgeSexFL.csv, or edit preprocess_data.py to produce the right .parquet files for training

3. make sure you have connection to transformer hub to download raceBERT/BERT models, or edit raceBERT_train.py to load models from local

Run

run the following commands in the root folder

python -m models.preprocess_data data/nmzpAgeSexFL.csv
python -m models.raceBERT_train raceBERT florida_5label

Details

preprocess_data.py

1. python -m models.preprocess_data [data_file]
2. if using different data, need to follow the column structure in nmzpAgeSexFL.csv

raceBERT_train.py

1. python -m models.raceBERT_train [model_label] [data_label]
2. check raceBERT_train.py for existing model_label and data_label, or edit raceBERT_train.py for custom options