Biknn

Code for Efficient Domain Adaptation for Non-Autoregressive Machine Translation

Requirements and Installation

• python >= 3.7
• pytorch >= 1.10.0
• faiss-gpu >= 1.7.3
• sacremoses == 0.0.41
• sacrebleu == 1.5.1
• fastBPE == 0.1.0
• scikit-learn >= 1.0.2
• seaborn >= 0.12.1
• editdistance >= 0.8.1
• elasticsearch >= 8.13.1

You can install this toolkit by

cd Biknn
pip install --editable ./

Note: Installing faiss with pip is not suggested. For stability, we recommand you to install faiss with conda

CPU version only:
conda install faiss-cpu -c pytorch

GPU version:
conda install faiss-gpu -c pytorch # For CUDA

Data

The data we used in the paper can be found here multi-domain de-en dataset

WMT19 data can be found wmt19

Get Datastore and Combiner

You can download the cached datastore and trained combiner at:

HuggingFace

# inference 
bash knnbox-scripts/inference.sh

You can download the datastore and pre-trained combiner and put them in the according dir, change the path in the script to your own path.

To generate the code, using the following command:

cd knnbox-scripts
# step 1. build datastore
bash build.sh
# step 2. renew datastore
bash renew.sh
# step 3. train metanetwork
bash train_metanetwork
# step 4. inference 
bash inference.sh

The code is primarily implemented through knn-mt and knnbox. We will release a cleaner version in future;