Download

wget http://aiweb.cs.ehime-u.ac.jp/~yamauchi/academic_model/Academic_BART_base.tar.gz

Hugging Face

Requirements

Preprocess

python ./apply-sp.py $TRAIN_SRC $DATASET_DIR/train.src-tgt -bpe_model $SENTENCEPIECE_MODELpython ./apply-sp.py $VALID_SRC $DATASET_DIR/valid.src-tgt -bpe_model $SENTENCEPIECE_MODELpython ./apply-sp.py $TEST_SRC $DATASET_DIR/test.src-tgt -bpe_model $SENTENCEPIECE_MODEL

fairseq-preprocess \    --source-lang "src" \    --target-lang "tgt" \    --trainpref "${DATASET_DIR}/train.src-tgt" \    --validpref "${DATASET_DIR}/valid.src-tgt" \    --testpref "${DATASET_DIR}/test.src-tgt" \    --destdir "data-bin/" \    --workers 60 \    --srcdict ${DICT} \    --tgtdict ${DICT}

Finetune

fairseq-train data-bin/ \    --restore-file $BART_PATH \    --max-tokens 512 --max-sentences $MAX_SENTENCES \    --task translation \    --source-lang src --target-lang tgt \    --truncate-source \    --layernorm-embedding \    --share-all-embeddings \    --share-decoder-input-output-embed \    --reset-optimizer --reset-dataloader --reset-meters \    --required-batch-size-multiple 1 \    --arch bart_base \    --criterion label_smoothed_cross_entropy \    --label-smoothing 0.1 \    --dropout 0.1 --attention-dropout 0.1 \    --weight-decay 0.01 --optimizer adam --adam-betas "(0.9, 0.999)" --adam-eps 1e-08 \    --clip-norm 0.1 \    --lr-scheduler polynomial_decay --lr $LR --total-num-update $TOTAL_NUM_UPDATES --warmup-updates $WARMUP_UPDATES \    --fp16 --update-freq $UPDATE_FREQ \    --skip-invalid-size-inputs-valid-test \    --no-epoch-checkpoints \    --save-interval-updates $SAVE_INTERVAL --save-dir result_test \    --patience 5 \

Reference