Empty conda environment

[csainzribera]

After following this part of the installation (see below) my conda env SE3nv has nothing installed in it (as shown by conda list when SE3nv is activated). If I try to run RFDiffusion in this conda env I get ModuleNotFoundError. What could be going wrong during the conda env setup?

conda env create -f env/SE3nv.yml

conda activate SE3nv cd env/SE3Transformer pip install --no-cache-dir -r requirements.txt python setup.py install cd ../.. # change into the root directory of the repository pip install -e . # install the rfdiffusion module from the root of the repository

Thanks!

[roccomoretti]

The conda env create -f env/SE3nv.yml should create the SE3nv environment and populate it with packages. If you're not seeing anything with a conda list, then something likely went wrong with that step. (Do you remember seeing any error messages?)

I'd probably recommend starting from scratch. Do a conda remove --name SE3nv --all to delete the environment (deactivate the environment first, if you have it active). Then repeat the installation instructions, paying close attention to the output of each command, looking for error messages that would indicate things did not run to completion.

[csainzribera]

Thank you so much for your response.

I have restarted from scratch and I indeed get an error:

`Found conflicts! Looking for incompatible packages. This can take several minutes. Press CTRL-C to abort. failed \

UnsatisfiableError: The following specifications were found to be incompatible with each other: ` The issue seems to be with __cuda and __glibc:

`- feature:/linux-64::__cuda==12.7=0

• feature:/linux-64::__glibc==2.35=0
• feature:|@/linux-64::__cuda==12.7=0
• feature:|@/linux-64::__glibc==2.35=0
• cudatoolkit=11.1 -> __glibc[version='>=2.17,<3.0.a0']
• cudatoolkit=11.1 -> libgcc-ng[version='>=10.3.0'] -> __glibc[version='>=2.17']
• dgl-cuda11.1 -> numpy -> __glibc[version='>=2.17,<3.0.a0']
• pip -> python[version='>=3.13,<3.14.0a0'] -> __glibc[version='>=2.17,<3.0.a0']
• python=3.9 -> __glibc[version='>=2.17,<3.0.a0']
• python=3.9 -> libgcc-ng[version='>=11.2.0'] -> __glibc[version='>=2.17']
• torchaudio -> __glibc[version='>=2.17,<3.0.a0']
• torchaudio -> libgcc-ng[version='>=11.2.0'] -> __glibc[version='>=2.17|>=2.28,<3.0.a0']
• torchaudio -> pytorch[version='>=2.5.1,<2.5.2.0a0'] -> __cuda[version='>=11.8|>=12.6|>=12.0']
• torchvision -> __cuda[version='>=11.8']
• torchvision -> __glibc[version='>=2.17|>=2.17,<3.0.a0|>=2.28,<3.0.a0']
• torchvision -> pytorch=2.3[build=cuda118] -> __cuda ` Thanks for your help.

[roccomoretti]

It looks specifically like the issue might be the torchvision and torchaudio dependencies, and that they're looking for a higher level of cuda for some reason.

See if editing the env/SE3nv.yml file to add the following version numbers helps (they're what I have for my working version):

- torchaudio=0.9.1
- torchvision=0.10.1

[csainzribera]

Thanks for your help. I no longer have access to the GPU I was using but will try this when I regain access.