README.md

First, activate the conda environment: conda activate m2qa_experiments
Then train the XLM-R-Base & XLM-R-Domain models by starting the bash scripts:

• XLM-R Base:
  • Execute xlm-r-fine-tune/train_full_xlm_r_on_squad.sh
• XLM-R Domain Adapted:
  1. Pretrain the XLM-R model on the target domain: Execute the bash files in the xlm-r-adapted directory
  2. Fine-tune each pretrained XLM-R model on SQuAD v2. The bash scripts to do this are in the Execute xlm-r-fine-tune directory
  • Example to train the XLM-R-Domain model for the creative writing domain:
    1. Execute xlm-r-adapted/pretrain_xlm_r_creative_writing.sh
    2. Execute xlm-r-fine-tune/train_xlm_r_pretrained_creative_writing.sh