Improving generalization of language models
Project inspired by Asa's work https://aclanthology.org/2021.mrl-1.20.pdf
Setup a conda env and install the below dependencies:
conda create --name nlp_gen
conda activate nlp_gen
conda install transformers
conda install -c pytorch pytorch
conda install -c anaconda scikit-learn
conda install -c conda-forge datasets
pip install pyhessian
To experiment with this repo:
- Clone this repository using:
git clone https://github.com/sakshambassi/nlp-generalization.git - Go inside the newly created dir:
cd nlp-generalization - Clone HFSharpness repo:
git clone https://github.com/sakshambassi/hfsharpness.git - Now run any experiment like:
python text_classification.py - or run it on Greene server:
sbatch greene_run.slurm