### Directory structure ###
├─config:
| There are several config.yaml files that include hyperparameters and model class name and several options.
|
├─data:
| Some data exist in this directory to use train or validation.
| This directory does not exist in gihub because of their size of storage, so you download this google drive.
|
├─origin_dataset
| ├─LARC_dataset
| ├─concept_eval
| ├─concept_total
| ├─concept_train
| ├─corpus
| ├─evaluation
| ├─training
| There exists original data files. If you can use preprocess files, then you can get data that is simliar data in google drive.
| LARC_dataset is text data - https://github.com/samacqua/LARC
| concept total/train/eval just are divide. So this data is included in corpus.
| corpus is ConceptARC dataset. That include 16 kinds ARC tasks and each kind involve 10 problems - https://github.com/victorvikram/ConceptARC
| training and evaluation are original ARC dataset.
|
├─dataset:
| Dataset class was defined in this directory
|
├─model:
| Model class was defined in this directory
|
├─preprocess:
| This directory have some code that used when data preprosessed.
|
├─trainer:
| This directory have some code that used when trained model.
|
├─utils:
| etc such as visualization or fix seed as so on.