conda env create -f environment.yml
# computation environments
kaldi
python>=3.8
cuda=11.0.3
pytorch=1.7.1Please refer to the installation details in https://github.com/matln/asv-subtools#ready-to-start.
git clone -b CDMA https://github.com/matln/asv-subtoolsChange to your own Kaldi & ASV-subtools path in path.sh.
export KALDI_ROOT=<your path>
export SUBTOOLS=<your path>It is recommended to change the stage and endstage to run step by step.
bash preprocess.sh- Pretraining
bash run_resnet34_pretrain.sh- Supervised domain adaptation
bash run_resnet34_SDA.sh- Unsupervised domain adaptation
bash run_resnet34_UDA.sh- Optimize the MMD loss in the embedding space
bash run_resnet34_emb_mmd.sh- Finetuning
bash run_resnet34_FT.sh