hf

Finetuning models

• run_sum.py is the main file for finetuning models on headline generation and abstractive summarization experiments for Varta and Xl-sum datasets.
• configs folder contains the configurations for training, model, and data.
• scripts folder contains the evaluation scripts to evaluate the model performance.

Note: all our experiments were run on pytorch compiled with Cuda 11.7

To run the experiments, activate the virtual environment and run the following line with the corresponding config file under configs folder.

python -m torch.distributed.launch --nproc_per_node [# of gpus] run_sum.py [config_file.json]