Primary config directory not found.

[dhiraj82]

Hi
I am trying to run rfdiffusion using docker image, but I am getting error that "Primary config directory not found." Why I am getting this error and how to fix this issue. I used following command to pull docker image.

apptainer pull rfd.sif docker://rosettacommons/rfdiffusion

Thanks
Dhiraj

[rclune]

Are you running the example shown on DockerHub? If not, can you share the full command that you are running that generates this issue and the full output?

[dhiraj82]

I am using this command to run rfdiffusion after starting apptainer in interactive mode.
apptainer shell -e ~/rfd.sif
run_inference.py inference.output_prefix=test inference.input_pdb=mypdeb.pdb inference.num_designs=10 'contigmap.contigs=[146-146/0 A1-211/0 C335-543]' diffuser.partial_T=20

Thanks
Dhiraj

[woodsh17]

Can you try pointing to run_inference.py inside the container when you are in interactive mode? Something like:
/app/RFdiffusion/scripts/run_inference.py inference.output_prefix=test inference.input_pdb=mypdeb.pdb inference.num_designs=10 'contigmap.contigs=[146-146/0 A1-211/0 C335-543]' diffuser.partial_T=20?

[dhiraj82]

I fixed this issue by making my own container and copying the config folder to location rfdiffusion is looking for. However now, using the instruction about rfdiffusion installation, it's installing pytorch cpu version and not gpu version. I tried specifically asking for pytorch-cuda but it's always installing cpu version.
I tried using rfdiffusion as base for my apptainer container but it's not finding numpy. How can I fix these problems?
Dhiraj

[rclune]

Is there a functionality that you need that can't be done without interactive mode? Does the RC-supported Docker image not work when used similar to the example syntax? (apptainer run --nv ...)

[dhiraj82]

I also tried apptainer run --nv ... but numpy version conflict is creating problem. I am getting this error -
ModuleNotFoundError: No module named 'numpy._core'
tensor flow need numpy 1.24 while jaxlib require 1.26.

Update:
I tried to install it on my new updated computer without apptainer and getting same error message- the full error mesage is -

Traceback (most recent call last):
File "/home/dsrivastava/RFdiffusion/scripts/run_inference.py", line 54, in main
sampler = iu.sampler_selector(conf)
File "/home/dsrivastava/RFdiffusion/rfdiffusion/inference/utils.py", line 511, in sampler_selector
sampler = model_runners.SelfConditioning(conf)
File "/home/dsrivastava/RFdiffusion/rfdiffusion/inference/model_runners.py", line 38, in init
self.initialize(conf)
File "/home/dsrivastava/RFdiffusion/rfdiffusion/inference/model_runners.py", line 148, in initialize
self.diffuser = Diffuser(**self._conf.diffuser, cache_dir=schedule_directory)
File "/home/dsrivastava/RFdiffusion/rfdiffusion/diffusion.py", line 582, in init
self.so3_diffuser = IGSO3(
File "/home/dsrivastava/RFdiffusion/rfdiffusion/diffusion.py", line 198, in init
self.igso3_vals = self._calc_igso3_vals(L=L)
File "/home/dsrivastava/RFdiffusion/rfdiffusion/diffusion.py", line 233, in _calc_igso3_vals
igso3_vals = read_pkl(cache_fname)
File "/home/dsrivastava/RFdiffusion/rfdiffusion/diffusion.py", line 144, in read_pkl
raise (e)
File "/home/dsrivastava/RFdiffusion/rfdiffusion/diffusion.py", line 140, in read_pkl
return pickle.load(handle)
ModuleNotFoundError: No module named 'numpy._core'

I tried binders design as well as partial diffusion with same error.

Thanks
Dhiraj