Grammatical acceptability evaluation using the material of the RuCoLA corpus

It is possible to open the entire project in Google.Colab or you can train and use the model locally.

Installation

pip install -r requirements.txt
./build_datasets.sh

Usage

1. Pre-train grammeme model

./pretrain.py

2. Train main model using one of the checkpoints of pre-trained model in data/model/

./main.py --checkpoint-dir data/model/<SOME_CHECKPOINT_DIR>