diff --git a/src/xminigrid/__init__.py b/src/xminigrid/__init__.py index 507be7c..f57ecbe 100644 --- a/src/xminigrid/__init__.py +++ b/src/xminigrid/__init__.py @@ -90,6 +90,14 @@ width=9, ) +register( + id="XLand-MiniGrid-R1-11x11", + entry_point="xminigrid.envs.xland:XLandMiniGrid", + grid_type="R1", + height=11, + width=11, +) + register( id="XLand-MiniGrid-R1-13x13", entry_point="xminigrid.envs.xland:XLandMiniGrid", @@ -98,6 +106,14 @@ width=13, ) +register( + id="XLand-MiniGrid-R1-15x15", + entry_point="xminigrid.envs.xland:XLandMiniGrid", + grid_type="R1", + height=15, + width=15, +) + register( id="XLand-MiniGrid-R1-17x17", entry_point="xminigrid.envs.xland:XLandMiniGrid", @@ -115,6 +131,14 @@ width=9, ) +register( + id="XLand-MiniGrid-R2-11x11", + entry_point="xminigrid.envs.xland:XLandMiniGrid", + grid_type="R2", + height=11, + width=11, +) + register( id="XLand-MiniGrid-R2-13x13", entry_point="xminigrid.envs.xland:XLandMiniGrid", @@ -123,6 +147,14 @@ width=13, ) +register( + id="XLand-MiniGrid-R2-15x15", + entry_point="xminigrid.envs.xland:XLandMiniGrid", + grid_type="R2", + height=15, + width=15, +) + register( id="XLand-MiniGrid-R2-17x17", entry_point="xminigrid.envs.xland:XLandMiniGrid", @@ -140,6 +172,14 @@ width=9, ) +register( + id="XLand-MiniGrid-R4-11x11", + entry_point="xminigrid.envs.xland:XLandMiniGrid", + grid_type="R4", + height=11, + width=11, +) + register( id="XLand-MiniGrid-R4-13x13", entry_point="xminigrid.envs.xland:XLandMiniGrid", @@ -148,6 +188,14 @@ width=13, ) +register( + id="XLand-MiniGrid-R4-15x15", + entry_point="xminigrid.envs.xland:XLandMiniGrid", + grid_type="R4", + height=15, + width=15, +) + register( id="XLand-MiniGrid-R4-17x17", entry_point="xminigrid.envs.xland:XLandMiniGrid", diff --git a/src/xminigrid/envs/xland.py b/src/xminigrid/envs/xland.py index 1cf7b82..ad26649 100644 --- a/src/xminigrid/envs/xland.py +++ b/src/xminigrid/envs/xland.py @@ -151,7 +151,7 @@ def default_params(self, **kwargs) -> XLandEnvParams: # this is just a heuristic to prevent brute force in one episode, # so that agent need to remember what he tried in previous episodes. # If this is too small, change it or increase number of trials (NB: these are not equivalent). - params = params.replace(max_steps=4 * (params.height * params.width)) + params = params.replace(max_steps=3 * (params.height * params.width)) return params def _generate_problem(self, params: XLandEnvParams, key: jax.Array) -> State[EnvCarry]: