Model Instantiator bugged if "concatenate" in output

[Xploeng]

Description

The model instantiators are bugged if you try to use "concatenate" in the output definition of a model. The documentation clearly states that this should be possible for inputs and outputs.
Below is an example for a failing instatiator.

source = gaussian_model(
        observation_space=obs_space,
        action_space=act_space,
        network=network,
        output="concatenate([head_a, head_b])",
        return_source=True,
    )

Minimal reproduction of the bug in skrl's model instantiator output parser:

skrl_concat_output_bug.py

Root cause: In _parse_output (common.py), the visit_Call method of the NodeTransformer replaces node.func with an ast.Attribute node when it encounters concatenate, but then falls through to the activation function check which assumes node.func is still an ast.Name (i.e. has .id).

Compare with _parse_input, which handles concatenate correctly by returning early before any activation logic.

Fix: Add else to the activation check only runs when the concatenate branch did not fire.

Error Log:

File "/home/jules/thesis/scripts/skrl/train.py", line 273, in main
    runner = Runner(env, agent_cfg)
             ^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jules/IsaacLab/lab_env/lib/python3.11/site-packages/skrl/utils/runner/torch/runner.py", line 33, in __init__
    self._models = self._generate_models(self._env, copy.deepcopy(self._cfg))
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jules/IsaacLab/lab_env/lib/python3.11/site-packages/skrl/utils/runner/torch/runner.py", line 247, in _generate_models
    source = model_class(
             ^^^^^^^^^^^^
  File "/home/jules/IsaacLab/lab_env/lib/python3.11/site-packages/skrl/utils/model_instantiators/torch/gaussian.py", line 77, in gaussian_model
    containers, output = generate_containers(network, output, embed_output=True, indent=1)
                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jules/IsaacLab/lab_env/lib/python3.11/site-packages/skrl/utils/model_instantiators/torch/common.py", line 288, in generate_containers
    output, output_modules, output_size = _parse_output(output)
                                          ^^^^^^^^^^^^^^^^^^^^^
  File "/home/jules/IsaacLab/lab_env/lib/python3.11/site-packages/skrl/utils/model_instantiators/torch/common.py", line 139, in _parse_output
    NodeTransformer().visit(tree)
  File "/home/jules/.local/share/uv/python/cpython-3.11.14-linux-x86_64-gnu/lib/python3.11/ast.py", line 418, in visit
    return visitor(node)
           ^^^^^^^^^^^^^
  File "/home/jules/.local/share/uv/python/cpython-3.11.14-linux-x86_64-gnu/lib/python3.11/ast.py", line 494, in generic_visit
    value = self.visit(value)
            ^^^^^^^^^^^^^^^^^
  File "/home/jules/.local/share/uv/python/cpython-3.11.14-linux-x86_64-gnu/lib/python3.11/ast.py", line 418, in visit
    return visitor(node)
           ^^^^^^^^^^^^^
  File "/home/jules/.local/share/uv/python/cpython-3.11.14-linux-x86_64-gnu/lib/python3.11/ast.py", line 503, in generic_visit
    new_node = self.visit(old_value)
               ^^^^^^^^^^^^^^^^^^^^^
  File "/home/jules/.local/share/uv/python/cpython-3.11.14-linux-x86_64-gnu/lib/python3.11/ast.py", line 418, in visit
    return visitor(node)
           ^^^^^^^^^^^^^
  File "/home/jules/IsaacLab/lab_env/lib/python3.11/site-packages/skrl/utils/model_instantiators/torch/common.py", line 121, in visit_Call
    activation = _get_activation_function(node.func.id, as_module=False)
                                          ^^^^^^^^^^^^
AttributeError: 'Attribute' object has no attribute 'id'

What skrl version are you using?

2.0.0

What ML framework/library version are you using?

PyTorch

Additional system information

Python 3.11, Ubuntu 22