acul3/distillbart-mnli

No Teacher BART distillation experiment for NLI tasks

DistilBart-MNLI

distillbart-mnli is the distilled version of bart-large-mnli created using the No Teacher Distillation technique proposed for BART summarisation by Huggingface, here.

We just copy alternating layers from bart-large-mnli and finetune more on the same data.

	matched acc	mismatched acc
bart-large-mnli (baseline, 12-12)	89.9	90.01
distilbart-mnli-12-1	87.08	87.5
distilbart-mnli-12-3	88.1	88.19
distilbart-mnli-12-6	89.19	89.01
distilbart-mnli-12-9	89.56	89.52

This is a very simple and effective technique, as we can see the performance drop is very little.

Fine-tuning

Clone and install transformers from source

git clone https://github.com/huggingface/transformers.git
pip install -qqq -U ./transformers

Download MNLI data

python transformers/utils/download_glue_data.py --data_dir glue_data --tasks MNLI

Create student model

python create_student.py \
  --teacher_model_name_or_path facebook/bart-large-mnli \
  --student_encoder_layers 12 \
  --student_decoder_layers 6 \
  --save_path student-bart-mnli-12-6 \

Start fine-tuning

python run_glue.py args.json

You can find the logs of these trained models in this wandb project.