Is conda a hard requirement?

[kuoirad]

I'm working with one of my users to get this installed, but conda is a non-starter here due to the licensing changes from a couple years back.

Pardon my relative inexperience with Python, but it looks like all you're doing with conda is using it for an environment. You're providing the S3-transform code and installing other stuff via pip. Is it possible this can be done with a "native" virtualenv instead?

[roccomoretti]

I haven't tried it, but from what I can tell, you should be able to manually handle the dependency management and skip Conda entirely -- I don't believe that there's anything which requires conda specifically.

[EricDeveaud]

conda is not required in order to install and run RFdiffusion.

just installed master branch (820bfdf) in a python-3.10 virtualenv

    # pytorch (adapt your version regarding cuda version)
    pip3 install torch==1.12.1+cu116 torchvision==0.13.1+cu116 torchaudio==0.12.1 --extra-index-url https://download.pytorch.org/whl/cu116
    # dgl (same adapt your version regarding cuda version)
    pip3 install dgl==1.0.2+cu116 -f https://data.dgl.ai/wheels/cu116/repo.html
    pip3 install e3nn==0.3.3 wandb==0.13.4 pynvml==11.0.0 hydra-core==1.3.2 pyrsistent==0.19.3
    pip3 install "git+https://github.com/NVIDIA/dllogger#egg=dllogger"
    cd RFdiffusion/env && pip3 install -r SE3Transformer/requirements.txt && pip3 install ./SE3Transformer
    # one module install numpy>=2 that breaks, revert to last 1.x version
    pip3 install numpy==1.26.4
    cd ../ && pip3 install  .

regards

Eric

[EricDeveaud]

You may want to install pyrosetta for make_secstruc_adj.py.

see https://graylab.jhu.edu/download/PyRosetta4/archive/release.
grab the wheel that fits your needs, and pip install pyrosetta.<FILL_IN_VERSION>.whl