Inconsistency in calculating product of observation/action space shapes

[songyuc]

Hi,

I noticed that there are two slightly different ways to calculate the product of the observation and action space shapes in examples/baselines/sac/sac.py.

Specifically, in the SoftQNetwork class, the following code is used:

nn.Linear(np.array(env.single_observation_space.shape).prod() + np.prod(env.single_action_space.shape), 256),

Here, np.array(env.single_observation_space.shape).prod() is used for the observation space, while np.prod(env.single_action_space.shape) is used for the action space.

It seems that both methods achieve the same result, but the first one (converting the tuple to a NumPy array) is slightly redundant, as np.prod() can directly handle tuples.

I'm curious about the reason for using these two different approaches. Is there a specific reason for this, or would it be beneficial to unify them for consistency?

Thank you for your time and effort in maintaining this project!