Skip to content

Latest commit

 

History

History
27 lines (23 loc) · 1.03 KB

README.md

File metadata and controls

27 lines (23 loc) · 1.03 KB
Raw

Abstraction with BART

Fine-tune BART directly on CNN/DM

python bart/finetune_bart.py

By default, model checkpoint will be saved to bart/checkpoints/checkpoint.pt. Only single GPU training is currently supported.

Fine-tune BART Paraphraser on BertSumExt outputs

python bart/finetune_bart.py --train-ext-output output/BertSumExt/score-1.01/bertsumext-out-train.txt --validation-ext-output output/BertSumExt/Score-0.5/bertsumext-out-validation.txt

Note that BertSumExt outputs on the training set may need to be generated first. By default, model checkpoint will be saved to bart/checkpoints/checkpoint.pt.

Evaluate on test set and dump output

python bart/score_bart.py --weight bart/checkpoints/best.pt --output-dir output/bertsumext-bart-out

Evaluate on test set output (ROUGE 1/2/L F1)

python bart/score_output.py --output-dir output/bertsumext-bart-out

The above commands should be run in the root directory. Use the commandline option -h for more information about script usage.