Customising or adapting CLM Script to work with GPU parallelism

I would like to customise or adapt an existing CLM Tuning script to work with GPU parallelism. I have heard that this can be done possibly with Deepespeed accelerator.config but am open to other options. Currently the script does not allow for GPU parallelism. Further links can be found here:

The script utilizes the datasets hosted on Bittensor's IPFS Genesis Dataset.

I would like to train the following models: https://huggingface.co/EleutherAI/gpt-neo-1.3B and https://huggingface.co/EleutherAI/gpt-neo-2.7B

The CLM Tuning script is here:
https://github.com/opentensor/clm_model_tuning
Details about it here:
https://docs.bittensor.com/nested/FineTuning.html
Bittensor dataset here:
https://docs.bittensor.com/nested/TheDataset.html