RFdiffusionRC

This repo provides usage instructions for setting up RFdiffusion on Penn State's Roar Collab (RC) HPC system.

Installation

1. under $PWD, make two dirs: env and model. Under env, mkdir SE3nv.
2. under the same dir as env and model, do git clone https://github.com/RosettaCommons/RFdiffusion.git
3. cd models and then do

wget http://files.ipd.uw.edu/pub/RFdiffusion/6f5902ac237024bdd0c176cb93063dc4/Base_ckpt.pt
wget http://files.ipd.uw.edu/pub/RFdiffusion/e29311f6f1bf1af907f9ef9f44b8328b/Complex_base_ckpt.pt
wget http://files.ipd.uw.edu/pub/RFdiffusion/60f09a193fb5e5ccdc4980417708dbab/Complex_Fold_base_ckpt.pt
wget http://files.ipd.uw.edu/pub/RFdiffusion/74f51cfb8b440f50d70878e05361d8f0/InpaintSeq_ckpt.pt
wget http://files.ipd.uw.edu/pub/RFdiffusion/76d00716416567174cdb7ca96e208296/InpaintSeq_Fold_ckpt.pt
wget http://files.ipd.uw.edu/pub/RFdiffusion/5532d2e1f3a4738decd58b19d633b3c3/ActiveSite_ckpt.pt
wget http://files.ipd.uw.edu/pub/RFdiffusion/12fc204edeae5b57713c5ad7dcb97d39/Base_epoch8_ckpt.pt

4. module load anaconda3/2021.05
5. cd ../env/SE3nv and conda create --prefix $PWD to install a local conda env
6. activate the installed conda env by source activate /storage/group/RISE/toShare/RFdiffusion/env/SE3nv
7. install pip locally by conda install pip
8. install dependencies for SE3Transformer using pip pip install jedi omegaconf hydra-core icecream pyrsistent
9. install dgl for SE3Transformer using pip pip install dgl==1.0.2+cu116 -f https://data.dgl.ai/wheels/cu116/repo.html
10. install SE3Transformer by going into RFdiffusion/env/SE3Transformer dir, cd ../../RFdiffusion/env/SE3Transformer/ and then pip -q install .
11. install additional dependencies for RFdiffusion: pip install torch opt_einsum e3nn
12. Go to the root dir of RFdiffusion repo using cd ../../ and then pip install . to install RFdiffusion

Testing

1. To make sure it's working, asking for a GPU or a CPU session by on RC
2. For some reason, the models should be in a specific directory: cp -r /storage/group/RISE/toShare/RFdiffusion/models /storage/group/RISE/toShare/RFdiffusion/env/SE3nv/lib/python3.11/site-packages/rfdiffusion/inference/../../models
3. Also, the input pdbs should be in a similar place cd /storage/group/RISE/toShare/RFdiffusion/env/SE3nv/lib/python3.11/site-packages/rfdiffusion/inference/../../ and mkdir examples
4. cd examples and cp -r /storage/group/RISE/toShare/RFdiffusion/RFdiffusion/examples/input_pdbs .
5. go to RFdiffusion/scripts and run ./run_inference.py 'contigmap.contigs=[150-150]' inference.output_prefix=test_outputs/test inference.num_designs=10

Links:

RFdiffusion github

RFdiffusion google colab notebook