Gym.SSourcediscrete ~seed id is gymnasium.make id with discrete actions. seed seeds the first reset, as env.reset(seed=seed); later resets continue from the environment's own generator, as in Gymnasium. Observations must be a Box; terminated and truncated are Gymnasium's. Raises Invalid_argument if the actions are not discrete. Raises Failure if Python cannot create the environment.
val discrete_with_renderer :
seed:int ->
string ->
(Owl.Neural.S.Algodiff.A.arr, int) Owl_rllib.Ndenv.t
* (unit ->
int * int * bytes)Like discrete, with render_mode="rgb_array". The callback returns width, height and row-major RGB bytes for the current state, after reset or step. Rendering does not advance the environment.
val continuous :
seed:int ->
string ->
(Owl.Neural.S.Algodiff.A.arr, Owl.Neural.S.Algodiff.A.arr) Owl_rllib.Ndenv.tThe same with Box actions.