Proper way to control the actuator via python?

[DrawZeroPoint]

Dear MuJoCo developers,

I am using the 2.1.4 version python binding of MuJoCo to control actuators with codes. So far, I encountered an odd issue that may be due to my improper way to use the API, so any suggestions are gratefully appreciated.

The simplified code I am using is like this:

actuator = self._data.actuator(actuator_name)
actuator.ctrl = cmd

where I can confirm that _data is initialized and contains the actuator; cmd is a plain float number. Oddly, the second line gives the error:

IndexError: too many indices for array: array is 0-dimensional, but 1 were indexed

I wonder if it is due to .ctrl should not be used in this way? What I planned to do is to assign the control value cmd to the actuator.
Previously with mujoco-py, I did the same thing like this:

self.sim.data.ctrl[actuator_id] = cmd

[JeanElsner]

I haven't really used the named accessors yet, but the ctrl field of the actuator view may not have been intended to be written into.
What I usually do is that I keep a list or map of indices that I initialise once like this:

id = mujoco.mj_name2id(model, mujoco.mjtObj.mjOBJ_ACTUATOR, name)

You can then assign values to the ctrl field like you used to, i.e.

data.ctrl[id] = ctrl_d

[saran-t]

You are using it correctly (or at least how I intended for it to be used). This is a bug in the Python bindings. I'm going to create an issue to track this bug, thanks for bringing it up!

[saran-t]

This should now be fixed in 55c1c91. You'll have to build it manually from source for now, though, until I get around to releasing a new version.

[saran-t]

The fixed has been incorporated into version 2.1.5 that was released today.

[DrawZeroPoint]

Thank you @saran-t for the quick fix! As your accepted answer, I followed Building from source, and here are some sidenotes I think might be helpful for revealing some details not addressed in the doc:

1. Inside the cloned ~/mujoco folder (a Python virtual environment needs to be activated to run make_sdist.sh):

cd python && ./make_sdist.sh

2. Then, set the env variable pointing to the binary of MuJoCo (here showcases my setup):

export MUJOCO_PATH=/home/dzp/.mujoco/mujoco214

3. Run pip wheel name_of_sdist.tar.gz, where name_of_sdist.tar.gz is generated in the last step inside mujoco/pyhon/dist

cd dist && pip wheel mujoco-2.1.4.tar.gz

4. Finally, install the wheel (you may need to prune the previous installation done by pip install mujoco):

pip install mujoco-2.1.4-cp38-cp38-linux_x86_64.whl

My work environment is Ubuntu 20.04 with Python 3.8. The aforementioned steps lead me successfully solved the issue.

Besides, I confirmed that @JeanElsner's circumvention is helpful for those who prefer an official release, thank you as well.