RFdiffusion
RFdiffusion copied to clipboard
pixi dependencies information for newbie.
I installed pixi and am leaving a message in case anyone needs information regarding dependencies.
[project]
name = "SE3nv"
version = "0.1.0"
channels = ["pytorch", "nvidia", "dglteam", "conda-forge", "main", "r", "msys2"]
platforms = ["linux-64"]
[tasks]
[dependencies]
python = "3.9.*"
pytorch = "2.0.*"
torchaudio = "*"
torchvision = "*"
cudatoolkit = "11.7.*"
"dgl-cuda11.7" = "*"
icecream = "*"
pip = "*"
mkl = "2024.0"
numpy = "1.26"
[pypi-dependencies]
hydra-core = "*"
pyrsistent = "*"
I proceeded with the installation using the following steps. Essentially, it's identical to the method provided in the repository.
# Begin with root of the repository
$pixi install
$pixi shell
$cd env/SE3Transformer
$pip install --no-cache-dir -r requirements.txt
$python setup.py install
$cd ../..
$pip install -e .
Results
❯ ./scripts/run_inference.py 'contigmap.contigs=[150-150]' inference.output_prefix=test_outputs/test inference.num_designs=10[2024-07-24 17:30:30,013][__main__][INFO] - Found GPU with device_name NVIDIA GeForce RTX 4090. Will run RFdiffusion on NVIDIA GeForce RTX 4090
Reading models from /home/ehottl/Downloads/RFdiffusion/rfdiffusion/inference/../../models
(skip)
[2024-07-24 17:32:56,674][rfdiffusion.inference.model_runners][INFO] - Timestep 2, input to next step: ------------------------------------------------------------------------------------------------------------------------------------------------------
[2024-07-24 17:32:57,543][__main__][INFO] - Finished design in 0.23 minutes
Hope, this will help someone.