Use a process for MuJoCo

[paul-nechifor]

No description provided.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[leshy]

I think ideally all this stuff is just part of the standard typed config for the mujoco module? not requesting to rewrite now, just mentioning

[paul-nechifor]

Ah, true, it should be configurable.

[jeff-hykin]

has the new scene_office1.xml been pushed to git lfs?

I checked out the branch, did a git lfs fetch/pull but it looks like load_model and model.py changed. Its now looking for data/mujoco_sim/scene_office1, which I can't seem to find/get. After checking git lfs I checked the sketchfab direct download but didn't see a good way to get it from there (I imagine the xml is more related to the robot anyways).

[jeff-hykin]

But good news: it works on MacOS if I merge this PR into mine and bypass that one issue (pickle the output from the old load_model and avoid calling load_model)!

Note, I did have to change the subprocess call to this:

            # mjpython must be used macOS (because of launch_passive inside mujoco_process.py)
            executable = sys.executable if sys.platform != 'darwin' else 'mjpython'
            self.process = subprocess.Popen(
                [executable, str(LAUNCHER_PATH), config_pickle, shm_names_json],
                stdout=subprocess.PIPE,
                stderr=subprocess.PIPE,
                text=True,
                bufsize=1,
            )

[paul-nechifor]

has the new scene_office1.xml been pushed to git lfs?

I checked out the branch, did a git lfs fetch/pull but it looks like load_model and model.py changed. Its now looking for data/mujoco_sim/scene_office1, which I can't seem to find/get. After checking git lfs I checked the sketchfab direct download but didn't see a good way to get it from there (I imagine the xml is more related to the robot anyways).

I'm guessing your data/mujoco_sim dir is old. It's extracted from data/.lfs/mujoco_sim.tar.gz. Just delete data/mujoco_sim and the .tar.gz file will be extracted again when you run again.

[leshy]

I don't care about details too much since you conform to Protocol for connection and you can own this, we won't depend on it for further components. Just one note on shared memory comms for the future

[leshy]

wasn't this whole file avoidable if we import & use standard shm or LCM pubsub?

pubsubs per spec have just .publish(topic, data) and .subscribe(topic, callback) so easy to use independantly. though I'm totally ok with this one shot implementation, it doesn't matter since we won't further build on top of this

[paul-nechifor]

I started like this because I thought it was easier to use without the callbacks. 😅 But, yeah, shouldn't be too different.