Question for issue like "'bool_' object cannot be converted to 'PyBool'"?

[dbsxdbsx]

·I am doing reinforcement learning with openAI gym from rust. When extracting information from rust after step, generally, almost all environments work fine, the core code like below:

 let py_array2 = array![action2].to_pyarray(self.py);
 let py_array3 = array![action3].to_pyarray(self.py);
 let py_array4 = array![action4].to_pyarray(self.py);

 let final_action: &Py<PyAny> = &[
                action1.to_object(self.py),
                vec![py_array2, py_array3, py_array4].to_object(self.py),
            ]
            .to_object(self.py).
            self.env
                .call_method1("step", (final_action,)) //not ok
                .expect("call step method failed")

...

let raw_obs = step_result
            .get_item(0)
            .expect("got observation failed from STEP");
let obs_vec = self.get_obs_vec_from_python(raw_obs);
let reward = step_result
    .get_item(1)
    .expect("get REWARD from step1 failed")
    .extract()
    .expect("get REWARD from step2 failed");
let is_done = step_result
    .get_item(2)
    .expect("get IS_DONE from step1 failed")
    .extract()
    .expect("get IS_DONE from step2 failed");// FAILED  for a specific environment

But for just a specific environment, it failed when code extracs is_done.
It gives error like:

PyErr { type: <class 'TypeError'>, value: TypeError("'bool_' object cannot be converted to 'PyBool'")

Here bool_ seems to be a boolean type from numpy? Anyway, it is strange, as the pure python version with this environment would just return a bool variable, nothing with numpy.

Do I miss something?

[davidhewitt]

I'm afraid that I don't have knowledge of your code, so don't know why a specific environment is producing numpy.bool_ values.

PyO3's rejection of this type is unsuprising, as numpy.bool_ does not inherit from bool:

>>> numpy.bool_.__mro__
(<class 'numpy.bool_'>, <class 'numpy.generic'>, <class 'object'>)
>>> bool.__mro__
(<class 'bool'>, <class 'int'>, <class 'object'>)
>>> issubclass(bool, int)
True
>>> issubclass(numpy.bool_, bool)
False

[dbsxdbsx]

@davidhewitt, thanks for your answer. I give it a deep try, and I found that the pure python version also output type of numpy.bool sometimes. And I just ingored it as everytime the numpy.bool object doesn't crash with code below:

if done:  # the numpy.bool__
    print(f"done in step {_step},with score:{episode_score}")
    # print("Episode finished after {} timesteps".format(t+1))
    break