embedding bunk beds for PLMs

To get started with EmBunkBed, you need to set up a conda environment and install the necessary dependencies. Follow the steps below to prepare your environment.

After cloning the repository, run

git submodule update --init --recursive

to checkout the desformers submodule.

• Anaconda3 (Version 2021.5)

First, load Anaconda3 and create a new conda environment named bunk_env with Python 3.11 and the CUDA Toolkit 11.3:

module load anaconda3/2021.5
conda create -n bunk_env python=3.11 cudatoolkit=11.3

Activate the newly created environment:

conda activate bunk_env

Navigate to the desformers directory:

cd desformers

Before installing the required packages, you need to modify the requirements.txt file:

1. Comment out any lines starting with "nvidia".
2. Comment out the line triton==2.1.0.

Now, you can install the remaining requirements using pip:

pip install -r requirements.txt

to-do: add to this section