update ModuRL

Cargo.toml

@@ -5,7 +5,7 @@ edition = "2024"
 
 [dependencies]
 candle-core = { git = "https://github.com/huggingface/candle", rev = "d205fb4" }
-modurl = { git = "https://github.com/ModuRL/ModuRL" }
+modurl = { git = "https://github.com/ModuRL/ModuRL", branch = "implement-determinism" }
 tracing = { version = "0.1.37", optional = true }
 log = { version = "0.4.28", optional = true }
 box2d-rs = "0.0.4"

src/box_2d/lunar_lander.rs

@@ -722,6 +722,7 @@ impl Default for LunarLanderV3 {
 
 impl Gym for LunarLanderV3 {
     type Error = candle_core::Error;
+    type SpaceError = candle_core::Error;
 
     fn reset(&mut self) -> Result<Tensor, Self::Error> {
         self.destroy();
@@ -1165,7 +1166,7 @@ impl Gym for LunarLanderV3 {
         })
     }
 
-    fn observation_space(&self) -> Box<dyn Space> {
+    fn observation_space(&self) -> Box<dyn Space<Error = Self::SpaceError>> {
         let low = vec![
             -2.5,
             -2.5,
@@ -1193,9 +1194,9 @@ impl Gym for LunarLanderV3 {
         Box::new(spaces::BoxSpace::new(low_tensor, high_tensor))
     }
 
-    fn action_space(&self) -> Box<dyn Space> {
+    fn action_space(&self) -> Box<dyn Space<Error = Self::SpaceError>> {
         // Discrete action space: 0=nop, 1=left engine, 2=main engine, 3=right engine
-        Box::new(spaces::Discrete::new(4, 0))
+        Box::new(spaces::Discrete::new(4))
     }
 }
 

src/classic_control/cartpole.rs

@@ -65,7 +65,7 @@ impl CartPoleV1 {
         let high = Tensor::from_vec(high, vec![4], device).expect("Failed to create tensor.");
         let low = Tensor::from_vec(low, vec![4], device).expect("Failed to create tensor.");
 
-        let action_space = spaces::Discrete::new(2, 0);
+        let action_space = spaces::Discrete::new(2);
         let observation_space = spaces::BoxSpace::new(low, high);
 
         Self {
@@ -233,6 +233,7 @@ impl Default for CartPoleV1 {
 
 impl Gym for CartPoleV1 {
     type Error = candle_core::Error;
+    type SpaceError = candle_core::Error;
 
     fn reset(&mut self) -> Result<Tensor, Self::Error> {
         self.steps_beyond_terminated = None;
@@ -346,11 +347,11 @@ impl Gym for CartPoleV1 {
         }
     }
 
-    fn observation_space(&self) -> Box<dyn Space> {
+    fn observation_space(&self) -> Box<dyn Space<Error = Self::SpaceError>> {
         Box::new(self.observation_space.clone())
     }
 
-    fn action_space(&self) -> Box<dyn Space> {
+    fn action_space(&self) -> Box<dyn Space<Error = Self::SpaceError>> {
         Box::new(self.action_space.clone())
     }
 }
@@ -457,7 +458,7 @@ mod tests {
         let _state = env.reset().expect("Failed to reset environment.");
         let action_space = env.action_space();
         for _ in 0..200 {
-            let action = action_space.sample(&Device::Cpu);
+            let action = action_space.sample(&Device::Cpu).unwrap();
             let StepInfo {
                 state: _,
                 reward: _,

src/classic_control/mountain_car.rs

@@ -44,7 +44,7 @@ impl MountainCarV0 {
         let low = Tensor::from_vec(low, vec![2], device).expect("Failed to create tensor.");
         let high = Tensor::from_vec(high, vec![2], device).expect("Failed to create tensor.");
 
-        let action_space = spaces::Discrete::new(3, 0);
+        let action_space = spaces::Discrete::new(3);
         let observation_space = spaces::BoxSpace::new(low, high);
 
         Self {
@@ -274,6 +274,7 @@ impl Default for MountainCarV0 {
 
 impl Gym for MountainCarV0 {
     type Error = candle_core::Error;
+    type SpaceError = candle_core::Error;
 
     fn reset(&mut self) -> Result<Tensor, Self::Error> {
         // Initialize position uniformly between -0.6 and -0.4
@@ -328,11 +329,11 @@ impl Gym for MountainCarV0 {
         })
     }
 
-    fn observation_space(&self) -> Box<dyn Space> {
+    fn observation_space(&self) -> Box<dyn Space<Error = Self::SpaceError>> {
         Box::new(self.observation_space.clone())
     }
 
-    fn action_space(&self) -> Box<dyn Space> {
+    fn action_space(&self) -> Box<dyn Space<Error = Self::SpaceError>> {
         Box::new(self.action_space.clone())
     }
 }
@@ -423,7 +424,7 @@ mod tests {
         env.reset().unwrap();
         let action_space = env.action_space();
         for _ in 0..200 {
-            let action = action_space.sample(&Device::Cpu);
+            let action = action_space.sample(&Device::Cpu).unwrap();
             let StepInfo {
                 state: _,
                 reward: _,