NVlabs / motion-policy-networks

This repo has the expert data generation infrastructure and Pytorch implementation of MPiNets.
MIT License
119 stars 19 forks source link

Update install instructions and config #2

Closed alaflaquiere-sony closed 1 year ago

alaflaquiere-sony commented 1 year ago

The installation instructions seem to be a bit out of date.

fishbotics commented 1 year ago

Thanks for making these changes @alaflaquiere-sony!

The only thing I'm wondering with this is about your change for the editable mode because I don't think it's necessary. If you notice in the docker run command in the documentation, it runs the command: 'export PYTHONPATH=/root/mpinets:$PYTHONPATH; git config --global --add safe.directory /root/mpinets; /bin/bash'. This should be placing this repo directly in your PYTHONPATH (so it should be importable).

Does that not work on your end? If it doesn't work without first installing it, we can definitely merge your change.

fishbotics commented 1 year ago

My other question: are you running the code in the docker or just on your host machine? If you're running it directly on your host machine, can you tell me what version of ROS you're using? The Docker runs Melodic and those franka_description paths were working when I tried them out in the Docker recently.

alaflaquiere-sony commented 1 year ago

Hi @fishbotics ,

I'm running in the docker you provided, and followed the instructions from the README. It doesn't seem the PYTHONPATH export works properly. Using your run command, I get the following PYTHONPATH inside the container:

# echo $PYTHONPATH 
# /opt/ros/melodic/lib/python2.7/dist-packages:/isaac-sim/exts/omni.isaac.lula/pip_prebundle/:/usr/lib/python3.7/site-packages

I think what happens is that the export only applies to the current bash, but another one is opened right after. Another solution would then be to append the export to the bashrc. This way, the export applies to any new bash that is spawned inside the container.

When did you download the franka_description package? Because afaik, they changed their xacro quite a while back (2021?). I had the same issue in a recent project of mine. This newer version is what I got building the docker image 2 days ago.

venishkrishna commented 1 year ago

is there any pre trained model or should we need to train it @fishbotics

fishbotics commented 1 year ago

Hey @venishkrishna , this is a better question to ask in an issue (so other people can find it). Can you please post it as an issue and then I can point you to the answer?