njunlp / selectivekd

Implementation for our AAAI-23 paper "Selective Knowledge Distillation for Non-Autoregressive Neural Machine Translation".

Selective KD

Implementation for our AAAI-23 paper "Selective Knowledge Distillation for Non-Autoregressive Neural Machine Translation" (arXiv).

Preparation

Train an autoregressive Transformer according to the instructions in Fairseq.

Use the trained autoregressive Transformer to generate target sentences for the training set.

You should make sure that the order of source sentences in distilled data matches that in raw data.

Then binarize raw and distilled data respectively.

python3 fairseq_cli/preprocess.py \
    --source-lang ${SRC} --target-lang ${TGT} \
    --trainpref ${INPUT_DIR}/train --validpref ${INPUT_DIR}/valid --testpref ${INPUT_DIR}/test \
    --destdir ${DATA_DIR} \
    --srcdict ${INPUT_DIR}/dict.${SRC}.txt --tgtdict ${INPUT_DIR}/dict.${TGT}.txt

Train an Evaluator

python3 fairseq_cli/train.py ${KD_DATA_DIR} \
    --arch glat_ctc --noise full_mask --share-all-embeddings \
    --criterion ctc_loss --label-smoothing 0.1 --lr 5e-4 --warmup-init-lr 1e-7 --stop-min-lr 1e-9 \
    --lr-scheduler inverse_sqrt --warmup-updates 4000 --optimizer adam --adam-betas '(0.9,0.999)' \
    --adam-eps 1e-6 --task translation_lev_modified --max-tokens 8192 --weight-decay 0.01 --dropout 0.1 \
    --encoder-layers 6 --encoder-embed-dim 512 --decoder-layers 6 --decoder-embed-dim 512 --fp16 \
    --max-source-positions 1000 --max-target-positions 1000 --max-update 100000 --seed 0 --clip-norm 2\
    --length-loss-factor 0 --log-interval 100 \
    --eval-bleu --eval-bleu-args '{"iter_decode_max_iter":0,"iter_decode_with_beam":1}' \
    --eval-tokenized-bleu --eval-bleu-remove-bpe --best-checkpoint-metric bleu \
    --maximize-best-checkpoint-metric --decoder-learned-pos --encoder-learned-pos \
    --apply-bert-init --activation-fn gelu --user-dir plugins --keep-best-checkpoints 5 \
    --save-dir ${EVALUATOR_DIR}

Train a Model Using Selective KD

python3 fairseq_cli/train.py ${RAW_DATA_DIR} --kd-data ${KD_DATA_DIR} \
    --arch glat_ctc --noise full_mask --share-all-embeddings --num-workers 4 \
    --criterion ctc_loss --label-smoothing 0.1 --lr 5e-4 --warmup-init-lr 1e-7 --stop-min-lr 1e-9 \
    --lr-scheduler inverse_sqrt --warmup-updates 4000 --optimizer adam --adam-betas '(0.9,0.999)' \
    --adam-eps 1e-6 --task translation_selective_kd --max-tokens 8192 --weight-decay 0.01 --dropout 0.1 \
    --encoder-layers 6 --encoder-embed-dim 512 --decoder-layers 6 --decoder-embed-dim 512 --fp16 \
    --max-source-positions 1000 --max-target-positions 1000 --max-update 300000 --seed 0 --clip-norm 2\
    --length-loss-factor 0 --log-interval 100 \
    --eval-bleu --eval-bleu-args '{"iter_decode_max_iter":0,"iter_decode_with_beam":1}' \
    --eval-tokenized-bleu --eval-bleu-remove-bpe --best-checkpoint-metric bleu \
    --maximize-best-checkpoint-metric --decoder-learned-pos --encoder-learned-pos \
    --apply-bert-init --activation-fn gelu --user-dir plugins \
    --keep-best-checkpoints 5 --save-dir ${MODEL_DIR} \
    --evaluator ${EVALUATOR_PATH} --load-from-checkpoint ${INITIALIZATION_MODEL} \
    --kd-threshold 0.4 --kd-threshold-offset 0.6

Inference

python3 fairseq_cli/generate.py ${RAW_DATA_DIR} \
    --user-dir plugins --gen-subset test \
    --task translation_selective_kd --remove-bpe --max-sentences 20 \
    --iter-decode-max-iter 0 --iter-decode-eos-penalty 0 --iter-decode-with-beam 1 \
    --path ${CHECKPOINT_PATH} --quiet

Citation

Please consider citing our paper if your find our research or this codebase helpful:

@article{liu2023selective,
  title={Selective Knowledge Distillation for Non-Autoregressive Neural Machine Translation},
  author={Liu, Min and Bao, Yu and Zhao, Chengqi and Huang, Shujian},
  journal={arXiv preprint arXiv:2303.17910},
  year={2023}
}

selectivekd's Issues

Could you please provide a script that installs all the additional packages?

I'm confussed about the environment, thanks a lot if you can help！