reset mask bug

[hrpan]

Minimal reproducing script:

    1 from ale_py.vector_env import AtariVectorEnv
    2 import numpy as np
    3 envs = AtariVectorEnv(game='breakout', num_envs=2)
    4 
    5 mask = np.array([False, True])
    6 envs.reset(options=dict(reset_mask=mask))

running this gives the following error:

Traceback (most recent call last):
  File "/home/hpan/project/tmp/bug.py", line 6, in <module>
    envs.reset(options=dict(reset_mask=mask))
  File "/home/hpan/project/pan/.venv/lib/python3.11/site-packages/ale_py/vector_env.py", line 170, in reset
    reset_indices, _ = np.where(reset_mask)
    ^^^^^^^^^^^^^^^^
ValueError: not enough values to unpack (expected 2, got 1)

I also tried calling envs.ale.reset directly, but then the script just hangs.
(something like envs.ale.reset(np.arange(1), np.arange(1)))

My goal is to implement the starting state decorrelation method described in https://arxiv.org/abs/1803.02811, which was found to stabilize training for parallel environments. Currently, I think there's no easy way to do this because step is always applied to all environments.

[pseudo-rnd-thoughts]

I don't get this error with the latest ale-py, but I do get the hanging
I'll investigate how we can fix this

[hrpan]

Just did a quick check and I think the np.where error is fixed in 66c19a4, but unfortunately is not yet included in the most recent release (0.11.2).