According to the above tutorial, I want to deploy openvla in my agx orin.
OpenVLA - NVIDIA Jetson AI Lab
But when I use the following instruction,
python3 -m mimicgen.generate
–tasks Stack_D4
–episodes 100
–output /data/datasets/mimicgen
–cameras agentview
–camera-width 224
–camera-height 224
an error occurs.
If MuJoCo is available in source code format, then you could build this yourself on ARM to produce the ARM version. If this is not available in source code form, then the providers of the software would have to build this for you.
Thank you. Can you give me some advice that how to build mujoco210
Where is their source code available? Basically, if you can compile this on a desktop PC running Ubuntu, then you would do the same thing directly on the Jetson. The AGX Xavier is quite capable of doing this, but you might end up needing more disk space, e.g., using a temporary USB external drive during the build. Building the source code though depends on what the source code is set up for.
An example of things to think about is that perhaps you need some particular library to run the program, and you might need a dev version of a library which contains just the headers for the functions in the library. A very common example is that libc is what is behind almost everything in any Linux or *NIX system, but when programs which use it are compiled, not only do they need libc, they also need the linux-libc-dev. libc is always installed, the dev package with headers which allows you to build your own is optional.
In some cases, for Ubuntu, the default package repositories will have whatever you need for adding build packages. In other cases you might need to add a repository in order to see the dev packages.
Also, it depends on the language used by the software. Maybe it is Python 2, maybe it is Python 3, in which case you don’t really compile it so much as you set it up. Much of what is out there is either in C or C++, and those are quite common to compile from source code. It isn’t possible to give much solid advice without knowing what the source looks like.
I have tried it. But the build of mujoco210 has failed.
Is there a forum or other support where you could post your build errors to for mujoco210? There might be something simple like a lack of some library.
mujoco210 doesn’t have an arm version. It only has the x86 version.
And I also tried to build it from source. More questions than I can track of.
I am assuming this is written in a compiled language; results will differ if this is something interpreted like Python.
The only way to proceed then is to build the source. I see in the original error screenshot that there are two things in particular which are issues:
/root/.mujoko/mujoko210. Is this supposed to be run as root?Do you know which library is in question?
Hi,
Which environment do you use? JetPack 5.1.x + AGX Xavier?
If so, do you pull the nano_llm container from the link below?
Thanks.