Democratizing Protein Language Models with Parameter-Efficient Fine-Tuning

🚧🚧 This is a pre-release 🚧🚧

Code from our manuscript on PEFT for protein language models.

Creating the environment using conda

Please use the provided env.yml file to install the required packages in a conda environment

conda env create -f env.yml

Running the code

Parameters to run the code, such as train and test mode, LoRA parameters (rank, matrices to adapt), data file paths, data splits etc. are provided in a configuration file. Sample configuration files are provided under ./ppi/config/ and ./symmetry/config

In addition to the configuration file, provide an identifier for your experiment as RUN_NAME and the gpu ids to use as DEVICES

Using run_job.sh

sh run_job_ppi.sh $RUN_NAME $CONFIG $DEVICES

sh run_job_symmetry.sh $RUN_NAME $CONFIG $DEVICES

Example: sh run_job.sh ppi_ft_expt ppi/config/config_ppi_finetune.yaml 0

Using python script

python ppi/main.py --run_name "$RUN_NAME" --config "$CONFIG" --devices "$DEVICES"