This is the implementation of Predicting the Validity of Set Data with Self-supervised Masked Transformer (SetMtr). We implement it (setmtr) based on a simple training toolkit torchility and provide a data preprocessing toolbox (./datasets)
pytorch>=2.0pytorch-lightning>=2.0,<2.1torchmetrics>=0.11,<0.12torchility == 0.9pip install torchility==0.9
-
Data Prepare
- run
datasets/original_data_process/<dataset_name>/data_gen.pyto generate data in the required format which will be saved indatasets/txts - run
datasets/process.pyto generate dataset for traning and evaluationg, which will bed saved indatasets/pkls.
- run
-
Configure
- Configure model parameters and data sets in
setmtr/config.yaml
- Configure model parameters and data sets in
-
Train and evaluate
python setmtr/train.py