AMP: Adversarial Mixing Policy for Relaxing Locally Linear Constraints in Mixup

This is the code for the paper "Adversarial Mixing Policy for Relaxing Locally Linear Constraints in Mixup" accepted at EMNLP'21

Install requirements

pip install -r requeriments.txt

Preparing data sets

Download link and unzip all the datasets into data fold.

Download pre-trained bert model and Glove embeddings

Create fold bert-base-uncased and enter the fold. Download $Bert_{base}$ model from hugging face. link

pytorch_model.bin
config.json
vocab.txt

Enter the project root directory. Download GloVe embeddings glove.840B.300d.zip from link

Run

Detailed descriptions of arguments are provided in run_main.py. For run the default parameters,

python run_main.py

License

MIT License