pyRDDLGym-jax 0.5__py3-none-any.whl → 1.0__py3-none-any.whl

pyRDDLGym_jax/examples/configs/Cartpole_Continuous_gym_drp.cfg

@@ -1,12 +1,12 @@
 [Model]
 logic='FuzzyLogic'
-logic_kwargs={'weight': 20}
-tnorm='ProductTNorm'
-tnorm_kwargs={}
+comparison_kwargs={'weight': 20}
+rounding_kwargs={'weight': 20}
+control_kwargs={'weight': 20}
 
 [Optimizer]
 method='JaxDeepReactivePolicy'
-method_kwargs={'topology': [32, 32]}
+method_kwargs={'topology': [32, 16]}
 optimizer='rmsprop'
 optimizer_kwargs={'learning_rate': 0.005}
 batch_size_train=1
@@ -14,5 +14,4 @@ batch_size_test=1
 
 [Training]
 key=42
-epochs=2000
-train_seconds=30
+epochs=1000

pyRDDLGym_jax/examples/configs/Cartpole_Continuous_gym_replan.cfg

@@ -1,8 +1,8 @@
 [Model]
 logic='FuzzyLogic'
-logic_kwargs={'weight': 100}
-tnorm='ProductTNorm'
-tnorm_kwargs={}
+comparison_kwargs={'weight': 50}
+rounding_kwargs={'weight': 50}
+control_kwargs={'weight': 50}
 
 [Optimizer]
 method='JaxStraightLinePlan'
@@ -15,6 +15,5 @@ rollout_horizon=30
 
 [Training]
 key=42
-epochs=1000
-train_seconds=1
+train_seconds=0.5
 print_summary=False

pyRDDLGym_jax/examples/configs/Cartpole_Continuous_gym_slp.cfg

@@ -1,19 +1,18 @@
 [Model]
 logic='FuzzyLogic'
-logic_kwargs={'weight': 20}
-tnorm='ProductTNorm'
-tnorm_kwargs={}
+comparison_kwargs={'weight': 30}
+rounding_kwargs={'weight': 30}
+control_kwargs={'weight': 30}
 
 [Optimizer]
 method='JaxStraightLinePlan'
 method_kwargs={}
 optimizer='rmsprop'
-optimizer_kwargs={'learning_rate': 0.001}
+optimizer_kwargs={'learning_rate': 0.002}
 batch_size_train=1
 batch_size_test=1
 clip_grad=1.0
 
 [Training]
 key=42
-epochs=5000
-train_seconds=30
+epochs=5000

pyRDDLGym_jax/examples/configs/HVAC_ippc2023_drp.cfg

@@ -1,8 +1,8 @@
 [Model]
 logic='FuzzyLogic'
-logic_kwargs={'weight': 10}
-tnorm='ProductTNorm'
-tnorm_kwargs={}
+comparison_kwargs={'weight': 5}
+rounding_kwargs={'weight': 5}
+control_kwargs={'weight': 5}
 
 [Optimizer]
 method='JaxDeepReactivePolicy'

pyRDDLGym_jax/examples/configs/HVAC_ippc2023_slp.cfg

@@ -1,14 +1,14 @@
 [Model]
 logic='FuzzyLogic'
-logic_kwargs={'weight': 10}
-tnorm='ProductTNorm'
-tnorm_kwargs={}
+comparison_kwargs={'weight': 5}
+rounding_kwargs={'weight': 5}
+control_kwargs={'weight': 5}
 
 [Optimizer]
 method='JaxStraightLinePlan'
 method_kwargs={}
 optimizer='rmsprop'
-optimizer_kwargs={'learning_rate': 0.01}
+optimizer_kwargs={'learning_rate': 0.02}
 batch_size_train=1
 batch_size_test=1
 

pyRDDLGym_jax/examples/configs/MountainCar_Continuous_gym_slp.cfg

@@ -1,8 +1,8 @@
 [Model]
 logic='FuzzyLogic'
-logic_kwargs={'weight': 10}
-tnorm='ProductTNorm'
-tnorm_kwargs={}
+comparison_kwargs={'weight': 10}
+rounding_kwargs={'weight': 10}
+control_kwargs={'weight': 10}
 
 [Optimizer]
 method='JaxStraightLinePlan'

pyRDDLGym_jax/examples/configs/MountainCar_ippc2023_slp.cfg

@@ -1,8 +1,8 @@
 [Model]
 logic='FuzzyLogic'
-logic_kwargs={'weight': 10}
-tnorm='ProductTNorm'
-tnorm_kwargs={}
+comparison_kwargs={'weight': 10}
+rounding_kwargs={'weight': 10}
+control_kwargs={'weight': 10}
 
 [Optimizer]
 method='JaxStraightLinePlan'

pyRDDLGym_jax/examples/configs/PowerGen_Continuous_drp.cfg

@@ -1,8 +1,8 @@
 [Model]
 logic='FuzzyLogic'
-logic_kwargs={'weight': 10}
-tnorm='ProductTNorm'
-tnorm_kwargs={}
+comparison_kwargs={'weight': 10}
+rounding_kwargs={'weight': 10}
+control_kwargs={'weight': 10}
 
 [Optimizer]
 method='JaxDeepReactivePolicy'

pyRDDLGym_jax/examples/configs/PowerGen_Continuous_replan.cfg

@@ -1,8 +1,8 @@
 [Model]
 logic='FuzzyLogic'
-logic_kwargs={'weight': 10}
-tnorm='ProductTNorm'
-tnorm_kwargs={}
+comparison_kwargs={'weight': 10}
+rounding_kwargs={'weight': 10}
+control_kwargs={'weight': 10}
 
 [Optimizer]
 method='JaxStraightLinePlan'

pyRDDLGym_jax/examples/configs/PowerGen_Continuous_slp.cfg

@@ -1,8 +1,8 @@
 [Model]
 logic='FuzzyLogic'
-logic_kwargs={'weight': 10}
-tnorm='ProductTNorm'
-tnorm_kwargs={}
+comparison_kwargs={'weight': 10}
+rounding_kwargs={'weight': 10}
+control_kwargs={'weight': 10}
 
 [Optimizer]
 method='JaxStraightLinePlan'

pyRDDLGym_jax/examples/configs/Quadcopter_drp.cfg

@@ -1,8 +1,8 @@
 [Model]
 logic='FuzzyLogic'
-logic_kwargs={'weight': 100}
-tnorm='ProductTNorm'
-tnorm_kwargs={}
+comparison_kwargs={'weight': 10}
+rounding_kwargs={'weight': 10}
+control_kwargs={'weight': 10}
 
 [Optimizer]
 method='JaxDeepReactivePolicy'

pyRDDLGym_jax/examples/configs/Quadcopter_slp.cfg

@@ -1,14 +1,14 @@
 [Model]
 logic='FuzzyLogic'
-logic_kwargs={'weight': 500}
-tnorm='ProductTNorm'
-tnorm_kwargs={}
+comparison_kwargs={'weight': 50}
+rounding_kwargs={'weight': 50}
+control_kwargs={'weight': 50}
 
 [Optimizer]
 method='JaxStraightLinePlan'
 method_kwargs={}
 optimizer='rmsprop'
-optimizer_kwargs={'learning_rate': 0.01}
+optimizer_kwargs={'learning_rate': 0.03}
 batch_size_train=1
 batch_size_test=1
 

pyRDDLGym_jax/examples/configs/Reservoir_Continuous_drp.cfg

@@ -1,8 +1,8 @@
 [Model]
 logic='FuzzyLogic'
-logic_kwargs={'weight': 10}
-tnorm='ProductTNorm'
-tnorm_kwargs={}
+comparison_kwargs={'weight': 10}
+rounding_kwargs={'weight': 10}
+control_kwargs={'weight': 10}
 
 [Optimizer]
 method='JaxDeepReactivePolicy'

pyRDDLGym_jax/examples/configs/Reservoir_Continuous_replan.cfg

@@ -1,8 +1,8 @@
 [Model]
 logic='FuzzyLogic'
-logic_kwargs={'weight': 20}
-tnorm='ProductTNorm'
-tnorm_kwargs={}
+comparison_kwargs={'weight': 10}
+rounding_kwargs={'weight': 10}
+control_kwargs={'weight': 10}
 
 [Optimizer]
 method='JaxStraightLinePlan'

pyRDDLGym_jax/examples/configs/Reservoir_Continuous_slp.cfg

@@ -1,8 +1,8 @@
 [Model]
 logic='FuzzyLogic'
-logic_kwargs={'weight': 10}
-tnorm='ProductTNorm'
-tnorm_kwargs={}
+comparison_kwargs={'weight': 10}
+rounding_kwargs={'weight': 10}
+control_kwargs={'weight': 10}
 
 [Optimizer]
 method='JaxStraightLinePlan'
@@ -14,5 +14,5 @@ batch_size_test=32
 
 [Training]
 key=42
-epochs=5000
-train_seconds=60
+epochs=2000
+train_seconds=30

pyRDDLGym_jax/examples/configs/UAV_Continuous_slp.cfg

@@ -1,14 +1,14 @@
 [Model]
 logic='FuzzyLogic'
-logic_kwargs={'weight': 1.0}
-tnorm='ProductTNorm'
-tnorm_kwargs={}
+comparison_kwargs={'weight': 1}
+rounding_kwargs={'weight': 1}
+control_kwargs={'weight': 1}
 
 [Optimizer]
 method='JaxStraightLinePlan'
 method_kwargs={}
 optimizer='rmsprop'
-optimizer_kwargs={'learning_rate': 0.0005}
+optimizer_kwargs={'learning_rate': 0.0003}
 batch_size_train=1
 batch_size_test=1
 

pyRDDLGym_jax/examples/configs/Wildfire_MDP_ippc2014_drp.cfg

@@ -1,8 +1,8 @@
 [Model]
 logic='FuzzyLogic'
-logic_kwargs={'weight': 100}
-tnorm='ProductTNorm'
-tnorm_kwargs={}
+comparison_kwargs={'weight': 100}
+rounding_kwargs={'weight': 100}
+control_kwargs={'weight': 100}
 
 [Optimizer]
 method='JaxDeepReactivePolicy'

pyRDDLGym_jax/examples/configs/Wildfire_MDP_ippc2014_replan.cfg

@@ -1,8 +1,8 @@
 [Model]
 logic='FuzzyLogic'
-logic_kwargs={'weight': 100}
-tnorm='ProductTNorm'
-tnorm_kwargs={}
+comparison_kwargs={'weight': 100}
+rounding_kwargs={'weight': 100}
+control_kwargs={'weight': 100}
 
 [Optimizer]
 method='JaxStraightLinePlan'

pyRDDLGym_jax/examples/configs/Wildfire_MDP_ippc2014_slp.cfg

@@ -1,8 +1,8 @@
 [Model]
 logic='FuzzyLogic'
-logic_kwargs={'weight': 100}
-tnorm='ProductTNorm'
-tnorm_kwargs={}
+comparison_kwargs={'weight': 100}
+rounding_kwargs={'weight': 100}
+control_kwargs={'weight': 100}
 
 [Optimizer]
 method='JaxStraightLinePlan'

pyRDDLGym_jax/examples/configs/default_drp.cfg

@@ -1,8 +1,8 @@
 [Model]
 logic='FuzzyLogic'
-logic_kwargs={'weight': 20}
-tnorm='ProductTNorm'
-tnorm_kwargs={}
+comparison_kwargs={'weight': 20}
+rounding_kwargs={'weight': 20}
+control_kwargs={'weight': 20}
 
 [Optimizer]
 method='JaxDeepReactivePolicy'

pyRDDLGym_jax/examples/configs/default_replan.cfg

@@ -1,8 +1,8 @@
 [Model]
 logic='FuzzyLogic'
-logic_kwargs={'weight': 20}
-tnorm='ProductTNorm'
-tnorm_kwargs={}
+comparison_kwargs={'weight': 20}
+rounding_kwargs={'weight': 20}
+control_kwargs={'weight': 20}
 
 [Optimizer]
 method='JaxStraightLinePlan'

pyRDDLGym_jax/examples/configs/default_slp.cfg

@@ -1,8 +1,8 @@
 [Model]
 logic='FuzzyLogic'
-logic_kwargs={'weight': 20}
-tnorm='ProductTNorm'
-tnorm_kwargs={}
+comparison_kwargs={'weight': 20}
+rounding_kwargs={'weight': 20}
+control_kwargs={'weight': 20}
 
 [Optimizer]
 method='JaxStraightLinePlan'

pyRDDLGym_jax/examples/configs/tuning_drp.cfg

@@ -0,0 +1,19 @@
+[Model]
+logic='FuzzyLogic'
+comparison_kwargs={'weight': MODEL_WEIGHT_TUNE}
+rounding_kwargs={'weight': MODEL_WEIGHT_TUNE}
+control_kwargs={'weight': MODEL_WEIGHT_TUNE}
+
+[Optimizer]
+method='JaxDeepReactivePolicy'
+method_kwargs={'topology': [LAYER1_TUNE, LAYER2_TUNE]}
+optimizer='rmsprop'
+optimizer_kwargs={'learning_rate': LEARNING_RATE_TUNE}
+batch_size_train=32
+batch_size_test=32
+
+[Training]
+train_seconds=30
+policy_hyperparams=POLICY_WEIGHT_TUNE
+print_summary=False
+print_progress=False

pyRDDLGym_jax/examples/configs/tuning_replan.cfg

@@ -0,0 +1,20 @@
+[Model]
+logic='FuzzyLogic'
+comparison_kwargs={'weight': MODEL_WEIGHT_TUNE}
+rounding_kwargs={'weight': MODEL_WEIGHT_TUNE}
+control_kwargs={'weight': MODEL_WEIGHT_TUNE}
+
+[Optimizer]
+method='JaxStraightLinePlan'
+method_kwargs={}
+optimizer='rmsprop'
+optimizer_kwargs={'learning_rate': LEARNING_RATE_TUNE}
+batch_size_train=32
+batch_size_test=32
+rollout_horizon=ROLLOUT_HORIZON_TUNE
+
+[Training]
+train_seconds=1
+policy_hyperparams=POLICY_WEIGHT_TUNE
+print_summary=False
+print_progress=False

pyRDDLGym_jax/examples/configs/tuning_slp.cfg

@@ -0,0 +1,19 @@
+[Model]
+logic='FuzzyLogic'
+comparison_kwargs={'weight': MODEL_WEIGHT_TUNE}
+rounding_kwargs={'weight': MODEL_WEIGHT_TUNE}
+control_kwargs={'weight': MODEL_WEIGHT_TUNE}
+
+[Optimizer]
+method='JaxStraightLinePlan'
+method_kwargs={}
+optimizer='rmsprop'
+optimizer_kwargs={'learning_rate': LEARNING_RATE_TUNE}
+batch_size_train=32
+batch_size_test=32
+
+[Training]
+train_seconds=30
+policy_hyperparams=POLICY_WEIGHT_TUNE
+print_summary=False
+print_progress=False

pyRDDLGym_jax/examples/run_plan.py

@@ -39,9 +39,12 @@ def main(domain, instance, method, episodes=1):
                       f'using default_{method}.cfg.', 'red')
         config_path = os.path.join(abs_path, 'configs', f'default_{method}.cfg') 
     planner_args, _, train_args = load_config(config_path)
+    if 'dashboard' in train_args: 
+        train_args['dashboard'].launch()
     
     # create the planning algorithm
-    planner = JaxBackpropPlanner(rddl=env.model, **planner_args)
+    planner = JaxBackpropPlanner(
+        rddl=env.model, dashboard_viz=env._visualizer, **planner_args)
     
     # evaluate the controller   
     if method == 'replan':

pyRDDLGym_jax/examples/run_tune.py

@@ -20,12 +20,19 @@ import os
 import sys
 
 import pyRDDLGym
-from pyRDDLGym.core.debug.exception import raise_warning
 
-from pyRDDLGym_jax.core.tuning import (
-    JaxParameterTuningDRP, JaxParameterTuningSLP, JaxParameterTuningSLPReplan
+from pyRDDLGym_jax.core.tuning import JaxParameterTuning, Hyperparameter
+from pyRDDLGym_jax.core.planner import (
+    load_config_from_string, JaxBackpropPlanner, 
+    JaxOfflineController, JaxOnlineController
 )
-from pyRDDLGym_jax.core.planner import load_config
+
+def power_2(x):
+    return int(2 ** x)
+
+
+def power_10(x):
+    return 10.0 ** x
 
 
 def main(domain, instance, method, trials=5, iters=20, workers=4):
@@ -35,31 +42,37 @@ def main(domain, instance, method, trials=5, iters=20, workers=4):
     
     # load the config file with planner settings
     abs_path = os.path.dirname(os.path.abspath(__file__))
-    config_path = os.path.join(abs_path, 'configs', f'{domain}_{method}.cfg')
-    if not os.path.isfile(config_path):
-        raise_warning(f'Config file {config_path} was not found, '
-                      f'using default_{method}.cfg.', 'red')
-        config_path = os.path.join(abs_path, 'configs', f'default_{method}.cfg') 
-    planner_args, plan_args, train_args = load_config(config_path)
+    config_path = os.path.join(abs_path, 'configs', f'tuning_{method}.cfg')
+    with open(config_path, 'r') as file:
+        config_template = file.read() 
+    
+    # map parameters in the config that will be tuned
+    hyperparams = [
+        Hyperparameter('MODEL_WEIGHT_TUNE', -1., 5., power_10),
+        Hyperparameter('POLICY_WEIGHT_TUNE', -2., 2., power_10),
+        Hyperparameter('LEARNING_RATE_TUNE', -5., 1., power_10),
+        Hyperparameter('LAYER1_TUNE', 1, 8, power_2),
+        Hyperparameter('LAYER2_TUNE', 1, 8, power_2),
+        Hyperparameter('ROLLOUT_HORIZON_TUNE', 1, min(env.horizon, 100), int)       
+    ]
     
-    # define algorithm to perform tuning
-    if method == 'slp':
-        tuning_class = JaxParameterTuningSLP        
-    elif method == 'drp':
-        tuning_class = JaxParameterTuningDRP    
-    elif method == 'replan':
-        tuning_class = JaxParameterTuningSLPReplan    
-    tuning = tuning_class(env=env,
-                          train_epochs=train_args['epochs'],
-                          timeout_training=train_args['train_seconds'],
-                          eval_trials=trials,
-                          planner_kwargs=planner_args,
-                          plan_kwargs=plan_args,
-                          num_workers=workers,
-                          gp_iters=iters)
+    # build the tuner and tune
+    tuning = JaxParameterTuning(env=env,
+                                config_template=config_template,
+                                hyperparams=hyperparams,
+                                online=method == 'replan',
+                                eval_trials=trials,
+                                num_workers=workers,
+                                gp_iters=iters)
+    tuning.tune(key=42, log_file=f'gp_{method}_{domain}_{instance}.csv')
     
-    # perform tuning and report best parameters
-    tuning.tune(key=train_args['key'], filename=f'gp_{method}', save_plot=True)
+    # evaluate the agent on the best parameters
+    planner_args, _, train_args = load_config_from_string(tuning.best_config)
+    planner = JaxBackpropPlanner(rddl=env.model, **planner_args)
+    klass = JaxOnlineController if method == 'replan' else JaxOfflineController
+    controller = klass(planner, **train_args)
+    controller.evaluate(env, episodes=1, verbose=True, render=True)
+    env.close()
 
 
 if __name__ == "__main__":