Closed swtyree closed 1 year ago
Hi @swtyree! I think 'sudo apt-get install liburdfdom-model' should help.
Thanks, @bibekyess. I am running Ubuntu 22.04, and the package you suggested doesn't appear to be available. Version 3.0 seems to be available (sudo apt-get install liburdfdom-model3.0
), but liburdfdom_model.so.3.0
isn't what Pinocchio is looking for. Do you have any other ideas?
This search would seem to indicate that it's only available for Ubuntu 18 and 20: https://pkgs.org/search/?q=liburdfdom_model.so.1.0
I use 'conda install conda-forge::urdfdom=1.0.4'.
I wonder which of these solutions seems more promising?
cosypose/lib3d/transform.py
to replace pinocchio
with another library, e.g. Transforms3Dpinocchio
claims to build/test successfully for Ubuntu 22.04I just saw this message from a few minutes ago. Let me try that first.
I use 'conda install conda-forge::urdfdom=1.0.4'.
@brieder, your suggestion was very helpful! After addressing the urdfdom
issue, I ran into an issue with an incompatibility between the specified pytorch/cudnn version and my system that runs CUDA 11.8. After fixing that, I am now able to run inference scripts, such as python -m cosypose.scripts.run_cosypose_eval --config tless-siso
.
Here are the changes that I needed to make to environment.yaml
:
urdfdom=1.0.4
to conda dependenciespytorch=1.3.1=py3.7_cuda10.1.243_cudnn7.6.3_0
and torchvision=0.4.2=py37_cu101
from conda dependenciestorch
and torchvision
to pip dependencies./deps/bullet3
with pybullet
(to address https://github.com/Simple-Robotics/cosypose/issues/2#issuecomment-1397890766)I've pasted the entire environment.yaml
file in this gist. Let me know if it would be helpful to have this as a pull request. I cannot confirm that these change work for all scripts, but I will reply here if I notice anything that doesn't work.
I just submitted a PR with my fixes for this issue (#11).
I am encountering the following error when attempting to run one of the eval scripts from the readme. I have run
sudo apt install liburdfdom-tools
andconda env create -n cosypose_SR --file environment.yaml
(though with./deps/bullet3
commented out due to the error mentioned in #2.