pyRDDLGym-jax 2.0__py3-none-any.whl → 2.2__py3-none-any.whl

pyRDDLGym_jax/core/visualization.py

@@ -20,8 +20,7 @@ import math
 import numpy as np
 import time
 import threading
-from typing import Any, Dict, List, Optional, Tuple, TYPE_CHECKING
-import warnings
+from typing import Any, Dict, Optional, Tuple, TYPE_CHECKING
 import webbrowser
 
 # prevent endless console prints
@@ -32,7 +31,7 @@ log.setLevel(logging.ERROR)
 import dash
 from dash.dcc import Interval, Graph, Store
 from dash.dependencies import Input, Output, State, ALL
-from dash.html import Div, B, H4, P, Img, Hr
+from dash.html import Div, B, H4, P, Hr
 import dash_bootstrap_components as dbc
 
 import plotly.colors as pc 
@@ -53,6 +52,7 @@ REWARD_ERROR_DIST_SUBPLOTS = 20
 MODEL_STATE_ERROR_HEIGHT = 300
 POLICY_STATE_VIZ_MAX_HEIGHT = 800
 GP_POSTERIOR_MAX_HEIGHT = 800
+GP_POSTERIOR_PIXELS = 100
 
 PLOT_AXES_FONT_SIZE = 11
 EXPERIMENT_ENTRY_FONT_SIZE = 14
@@ -1417,7 +1417,7 @@ class JaxPlannerDashboard:
         self.pgpe_return[experiment_id].append(callback['pgpe_return'])
         
         # data for return distributions
-        progress = callback['progress']
+        progress = int(callback['progress'])
         if progress - self.return_dist_last_progress[experiment_id] \
             >= PROGRESS_FOR_NEXT_RETURN_DIST:
             self.return_dist_ticks[experiment_id].append(iteration)
@@ -1486,8 +1486,8 @@ class JaxPlannerDashboard:
                 if i2 > i1:
                     
                     # Generate a grid for visualization
-                    p1_values = np.linspace(*bounds[param1], 100)
-                    p2_values = np.linspace(*bounds[param2], 100)
+                    p1_values = np.linspace(*bounds[param1], GP_POSTERIOR_PIXELS)
+                    p2_values = np.linspace(*bounds[param2], GP_POSTERIOR_PIXELS)
                     P1, P2 = np.meshgrid(p1_values, p2_values)
                     
                     # Predict the mean and deviation of the surrogate model
@@ -1500,8 +1500,7 @@ class JaxPlannerDashboard:
                     for p1, p2 in zip(np.ravel(P1), np.ravel(P2)):
                         params = {param1: p1, param2: p2}
                         params.update(fixed_params)
-                        param_grid.append(
-                            [params[key] for key in optimizer.space.keys])
+                        param_grid.append([params[key] for key in optimizer.space.keys])
                     param_grid = np.asarray(param_grid)
                     mean, std = optimizer._gp.predict(param_grid, return_std=True)
                     mean = mean.reshape(P1.shape)

pyRDDLGym_jax/examples/run_tune.py

@@ -3,7 +3,7 @@ is performed using a batched parallelized Bayesian optimization.
 
 The syntax is:
 
-    python run_tune.py <domain> <instance> <method> [<trials>] [<iters>] [<workers>]
+    python run_tune.py <domain> <instance> <method> [<trials>] [<iters>] [<workers>] [<dashboard>]
     
 where:
     <domain> is the name of a domain located in the /Examples directory
@@ -15,6 +15,7 @@ where:
     (defaults to 20)
     <workers> is the number of parallel workers (i.e. batch size), which must
     not exceed the number of cores available on the machine (defaults to 4)
+    <dashboard> is whether the dashboard is displayed
 '''
 import os
 import sys
@@ -35,7 +36,7 @@ def power_10(x):
     return 10.0 ** x
 
 
-def main(domain, instance, method, trials=5, iters=20, workers=4):
+def main(domain, instance, method, trials=5, iters=20, workers=4, dashboard=False):
     
     # set up the environment   
     env = pyRDDLGym.make(domain, instance, vectorized=True)
@@ -48,9 +49,9 @@ def main(domain, instance, method, trials=5, iters=20, workers=4):
     
     # map parameters in the config that will be tuned
     hyperparams = [
-        Hyperparameter('MODEL_WEIGHT_TUNE', -1., 5., power_10),
+        Hyperparameter('MODEL_WEIGHT_TUNE', -1., 4., power_10),
         Hyperparameter('POLICY_WEIGHT_TUNE', -2., 2., power_10),
-        Hyperparameter('LEARNING_RATE_TUNE', -5., 1., power_10),
+        Hyperparameter('LEARNING_RATE_TUNE', -5., 0., 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)       
@@ -64,7 +65,9 @@ def main(domain, instance, method, trials=5, iters=20, workers=4):
                                 eval_trials=trials,
                                 num_workers=workers,
                                 gp_iters=iters)
-    tuning.tune(key=42, log_file=f'gp_{method}_{domain}_{instance}.csv')
+    tuning.tune(key=42, 
+                log_file=f'gp_{method}_{domain}_{instance}.csv', 
+                show_dashboard=dashboard)
     
     # evaluate the agent on the best parameters
     planner_args, _, train_args = load_config_from_string(tuning.best_config)
@@ -77,7 +80,7 @@ def main(domain, instance, method, trials=5, iters=20, workers=4):
 
 def run_from_args(args):
     if len(args) < 3:
-        print('python run_tune.py <domain> <instance> <method> [<trials>] [<iters>] [<workers>]')
+        print('python run_tune.py <domain> <instance> <method> [<trials>] [<iters>] [<workers>] [<dashboard>]')
         exit(1)
     if args[2] not in ['drp', 'slp', 'replan']:
         print('<method> in [drp, slp, replan]')
@@ -86,6 +89,7 @@ def run_from_args(args):
     if len(args) >= 4: kwargs['trials'] = int(args[3])
     if len(args) >= 5: kwargs['iters'] = int(args[4])
     if len(args) >= 6: kwargs['workers'] = int(args[5])
+    if len(args) >= 7: kwargs['dashboard'] = bool(args[6])
     main(**kwargs) 
 
 

{pyRDDLGym_jax-2.0.dist-info → pyrddlgym_jax-2.2.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: pyRDDLGym-jax
-Version: 2.0
+Version: 2.2
 Summary: pyRDDLGym-jax: automatic differentiation for solving sequential planning problems in JAX.
 Home-page: https://github.com/pyrddlgym-project/pyRDDLGym-jax
 Author: Michael Gimelfarb, Ayal Taitler, Scott Sanner
@@ -58,18 +58,21 @@ Dynamic: summary
 
 Purpose:
 
-1. automatic translation of any RDDL description file into a differentiable simulator in JAX
-2. flexible policy class representations, automatic model relaxations for working in discrete and hybrid domains, and Bayesian hyper-parameter tuning.
+1. automatic translation of RDDL description files into differentiable JAX simulators
+2. implementation of (highly configurable) operator relaxations for working in discrete and hybrid domains
+3. flexible policy representations and automated Bayesian hyper-parameter tuning
+4. interactive dashboard for dyanmic visualization and debugging
+5. hybridization with parameter-exploring policy gradients.
 
 Some demos of solved problems by JaxPlan:
 
 <p align="middle">
-<img src="Images/intruders.gif" width="120" height="120" margin=0/>
-<img src="Images/marsrover.gif" width="120" height="120" margin=0/>
-<img src="Images/pong.gif" width="120" height="120" margin=0/>
-<img src="Images/quadcopter.gif" width="120" height="120" margin=0/>
-<img src="Images/reacher.gif" width="120" height="120" margin=0/>
-<img src="Images/reservoir.gif" width="120" height="120" margin=0/>
+<img src="https://github.com/pyrddlgym-project/pyRDDLGym-jax/blob/main/Images/intruders.gif" width="120" height="120" margin=0/>
+<img src="https://github.com/pyrddlgym-project/pyRDDLGym-jax/blob/main/Images/marsrover.gif" width="120" height="120" margin=0/>
+<img src="https://github.com/pyrddlgym-project/pyRDDLGym-jax/blob/main/Images/pong.gif" width="120" height="120" margin=0/>
+<img src="https://github.com/pyrddlgym-project/pyRDDLGym-jax/blob/main/Images/quadcopter.gif" width="120" height="120" margin=0/>
+<img src="https://github.com/pyrddlgym-project/pyRDDLGym-jax/blob/main/Images/reacher.gif" width="120" height="120" margin=0/>
+<img src="https://github.com/pyrddlgym-project/pyRDDLGym-jax/blob/main/Images/reservoir.gif" width="120" height="120" margin=0/>
 </p>
 
 > [!WARNING]  
@@ -219,7 +222,7 @@ Since version 1.0, JaxPlan has an optional dashboard that allows keeping track o
 and visualization of the policy or model, and other useful debugging features.
 
 <p align="middle">
-<img src="Images/dashboard.png" width="480" height="248" margin=0/>
+<img src="https://github.com/pyrddlgym-project/pyRDDLGym-jax/blob/main/Images/dashboard.png" width="480" height="248" margin=0/>
 </p>
 
 To run the dashboard, add the following entry to your config file:
@@ -235,8 +238,23 @@ More documentation about this and other new features will be coming soon.
 
 ## Tuning the Planner
 
-It is easy to tune the planner's hyper-parameters efficiently and automatically using Bayesian optimization.
-To do this, first create a config file template with patterns replacing concrete parameter values that you want to tune, e.g.:
+A basic run script is provided to run automatic Bayesian hyper-parameter tuning for the most sensitive parameters of JaxPlan:
+
+```shell
+jaxplan tune <domain> <instance> <method> <trials> <iters> <workers> <dashboard>
+```
+
+where:
+- ``domain`` is the domain identifier as specified in rddlrepository
+- ``instance`` is the instance identifier
+- ``method`` is the planning method to use (i.e. drp, slp, replan)
+- ``trials`` is the (optional) number of trials/episodes to average in evaluating each hyper-parameter setting
+- ``iters`` is the (optional) maximum number of iterations/evaluations of Bayesian optimization to perform
+- ``workers`` is the (optional) number of parallel evaluations to be done at each iteration, e.g. the total evaluations = ``iters * workers``
+- ``dashboard`` is whether the optimizations are tracked in the dashboard application.
+
+It is easy to tune a custom range of the planner's hyper-parameters efficiently. 
+First create a config file template with patterns replacing concrete parameter values that you want to tune, e.g.:
 
 ```ini
 [Model]
@@ -260,7 +278,7 @@ train_on_reset=True
 
 would allow to tune the sharpness of model relaxations, and the learning rate of the optimizer.
 
-Next, you must link the patterns in the config with concrete hyper-parameter ranges the tuner will understand:
+Next, you must link the patterns in the config with concrete hyper-parameter ranges the tuner will understand, and run the optimizer:
 
 ```python
 import pyRDDLGym
@@ -292,21 +310,7 @@ tuning = JaxParameterTuning(env=env,
                             gp_iters=iters)
 tuning.tune(key=42, log_file='path/to/log.csv')
 ```
- 
-A basic run script is provided to run the automatic hyper-parameter tuning for the most sensitive parameters of JaxPlan:
-
-```shell
-jaxplan tune <domain> <instance> <method> <trials> <iters> <workers>
-```
-
-where:
-- ``domain`` is the domain identifier as specified in rddlrepository
-- ``instance`` is the instance identifier
-- ``method`` is the planning method to use (i.e. drp, slp, replan)
-- ``trials`` is the (optional) number of trials/episodes to average in evaluating each hyper-parameter setting
-- ``iters`` is the (optional) maximum number of iterations/evaluations of Bayesian optimization to perform
-- ``workers`` is the (optional) number of parallel evaluations to be done at each iteration, e.g. the total evaluations = ``iters * workers``.
- 
+  
 
 ## Simulation
 
@@ -344,7 +348,16 @@ The [following citation](https://ojs.aaai.org/index.php/ICAPS/article/view/31480
 ```
 
 Some of the implementation details derive from the following literature, which you may wish to also cite in your research papers:
-- [A Distributional Framework for Risk-Sensitive End-to-End Planning in Continuous MDPs](https://ojs.aaai.org/index.php/AAAI/article/view/21226)
+- [A Distributional Framework for Risk-Sensitive End-to-End Planning in Continuous MDPs, AAAI 2022](https://ojs.aaai.org/index.php/AAAI/article/view/21226)
 - [Deep reactive policies for planning in stochastic nonlinear domains, AAAI 2019](https://ojs.aaai.org/index.php/AAAI/article/view/4744)
+- [Stochastic Planning with Lifted Symbolic Trajectory Optimization, AAAI 2019](https://ojs.aaai.org/index.php/ICAPS/article/view/3467/3335)
 - [Scalable planning with tensorflow for hybrid nonlinear domains, NeurIPS 2017](https://proceedings.neurips.cc/paper/2017/file/98b17f068d5d9b7668e19fb8ae470841-Paper.pdf)
-
+- [Baseline-Free Sampling in Parameter Exploring Policy Gradients: Super Symmetric PGPE, ANN 2015](https://link.springer.com/chapter/10.1007/978-3-319-09903-3_13)
+
+The model relaxations in JaxPlan are based on the following works:
+- [Poisson Variational Autoencoder, NeurIPS 2025](https://proceedings.neurips.cc/paper_files/paper/2024/file/4f3cb9576dc99d62b80726690453716f-Paper-Conference.pdf)
+- [Analyzing Differentiable Fuzzy Logic Operators, AI 2022](https://www.sciencedirect.com/science/article/pii/S0004370221001533)
+- [Learning with algorithmic supervision via continuous relaxations, NeurIPS 2021](https://proceedings.neurips.cc/paper_files/paper/2021/file/89ae0fe22c47d374bc9350ef99e01685-Paper.pdf)
+- [Universally quantized neural compression, NeurIPS 2020](https://papers.nips.cc/paper_files/paper/2020/file/92049debbe566ca5782a3045cf300a3c-Paper.pdf)
+- [Generalized Gumbel-Softmax Gradient Estimator for Generic Discrete Random Variables, 2020](https://arxiv.org/pdf/2003.01847)
+- [Categorical Reparametrization with Gumbel-Softmax, ICLR 2017](https://openreview.net/pdf?id=rkE3y85ee)

{pyRDDLGym_jax-2.0.dist-info → pyrddlgym_jax-2.2.dist-info}/RECORD

@@ -1,12 +1,12 @@
-pyRDDLGym_jax/__init__.py,sha256=TiPG4w8nN4AzPkhugwVvZkHmAgP955NltD4QRmBLhRU,19
+pyRDDLGym_jax/__init__.py,sha256=lqo7WXKfZGHPIOxgE6EWI5fGZHP2h6XrwVNNVQAUN3Q,19
 pyRDDLGym_jax/entry_point.py,sha256=dxDlO_5gneEEViwkLCg30Z-KVzUgdRXaKuFjoZklkA0,974
 pyRDDLGym_jax/core/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-pyRDDLGym_jax/core/compiler.py,sha256=Rn-aIqfgfWqu45bvCfPb9tB8RIOBVdbj-pI-V3WS2Z8,89212
-pyRDDLGym_jax/core/logic.py,sha256=_A6eGYtLVU3pbLAezxJVB9bnClJoaFIa2mBIDdFrqoU,39655
-pyRDDLGym_jax/core/planner.py,sha256=4j56l7SL7F89g2QA4nOpyhODmY0DamvxYLfCMKxJNbQ,118593
+pyRDDLGym_jax/core/compiler.py,sha256=_ERueJW7GQ7S8-IezreeuLs3fNCZbQZ8j7VMUVlEt1k,82306
+pyRDDLGym_jax/core/logic.py,sha256=ZeCwCLqC6BvXpRT06TvE2bfPNO6ALuMzPmUvXNzW6Uw,52278
+pyRDDLGym_jax/core/planner.py,sha256=0rluBXKGNHRPEPfegOWcx9__cJHr8KjZdDJtG7i1JjI,122793
 pyRDDLGym_jax/core/simulator.py,sha256=DnPL93WVCMZqtqMUoiJdfWcH9pEvNgGfDfO4NV0wIS0,9271
 pyRDDLGym_jax/core/tuning.py,sha256=RKKtDZp7unvfbhZEoaunZtcAn5xtzGYqXBB_Ij_Aapc,24205
-pyRDDLGym_jax/core/visualization.py,sha256=XtQL1A5dQIlfeUpte-r3lNVw-GNLxj2EYUNMz7AFOtc,70359
+pyRDDLGym_jax/core/visualization.py,sha256=4BghMp8N7qtF0tdyDSqtxAxNfP9HPrQWTiXzAMJmx7o,70365
 pyRDDLGym_jax/core/assets/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 pyRDDLGym_jax/core/assets/favicon.ico,sha256=RMMrI9YvmF81TgYG7FO7UAre6WmYFkV3B2GmbA1l0kM,175085
 pyRDDLGym_jax/examples/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -14,7 +14,7 @@ pyRDDLGym_jax/examples/run_gradient.py,sha256=KhXvijRDZ4V7N8NOI2WV8ePGpPna5_vnET
 pyRDDLGym_jax/examples/run_gym.py,sha256=rXvNWkxe4jHllvbvU_EOMji_2-2k5d4tbBKhpMm_Gaw,1526
 pyRDDLGym_jax/examples/run_plan.py,sha256=v2AvwgIa4Ejr626vBOgWFJIQvay3IPKWno02ztIFCYc,2768
 pyRDDLGym_jax/examples/run_scipy.py,sha256=wvcpWCvdjvYHntO95a7JYfY2fuCMUTKnqjJikW0PnL4,2291
-pyRDDLGym_jax/examples/run_tune.py,sha256=zqrhvLR5PeWJv0NsRxDCzAPmvgPgz_1NrtM1xBy6ndU,3606
+pyRDDLGym_jax/examples/run_tune.py,sha256=WbGO8RudIK-cPMAMKvI8NbFQAqkG-Blbnta3Efsep6c,3828
 pyRDDLGym_jax/examples/configs/Cartpole_Continuous_gym_drp.cfg,sha256=mE8MqhOlkHeXIGEVrnR3QY6I-_iy4uxFYRA71P1bmtk,347
 pyRDDLGym_jax/examples/configs/Cartpole_Continuous_gym_replan.cfg,sha256=nFFYHCKQUMn8x-OpJwu2pwe1tycNSJ8iAIwSkCBn33E,370
 pyRDDLGym_jax/examples/configs/Cartpole_Continuous_gym_slp.cfg,sha256=eJ3HvHjODoKdtX7u-AM51xQaHJnYgzEy2t3omNG2oCs,340
@@ -41,9 +41,9 @@ pyRDDLGym_jax/examples/configs/default_slp.cfg,sha256=mJo0woDevhQCSQfJg30ULVy9qG
 pyRDDLGym_jax/examples/configs/tuning_drp.cfg,sha256=CQMpSCKTkGioO7U82mHMsYWFRsutULx0V6Wrl3YzV2U,504
 pyRDDLGym_jax/examples/configs/tuning_replan.cfg,sha256=m_0nozFg_GVld0tGv92Xao_KONFJDq_vtiJKt5isqI8,501
 pyRDDLGym_jax/examples/configs/tuning_slp.cfg,sha256=KHu8II6CA-h_HblwvWHylNRjSvvGS3VHxN7JQNR4p_Q,464
-pyRDDLGym_jax-2.0.dist-info/LICENSE,sha256=Y0Gi6H6mLOKN-oIKGZulQkoTJyPZeAaeuZu7FXH-meg,1095
-pyRDDLGym_jax-2.0.dist-info/METADATA,sha256=ZYIe9c_Tar4WO8qQOvcUIJVMmZznPUBRaegS0DH2un8,15090
-pyRDDLGym_jax-2.0.dist-info/WHEEL,sha256=In9FTNxeP60KnTkGw7wk6mJPYd_dQSjEZmXdBdMCI-8,91
-pyRDDLGym_jax-2.0.dist-info/entry_points.txt,sha256=Q--z9QzqDBz1xjswPZ87PU-pib-WPXx44hUWAFoBGBA,59
-pyRDDLGym_jax-2.0.dist-info/top_level.txt,sha256=n_oWkP_BoZK0VofvPKKmBZ3NPk86WFNvLhi1BktCbVQ,14
-pyRDDLGym_jax-2.0.dist-info/RECORD,,
+pyrddlgym_jax-2.2.dist-info/LICENSE,sha256=Y0Gi6H6mLOKN-oIKGZulQkoTJyPZeAaeuZu7FXH-meg,1095
+pyrddlgym_jax-2.2.dist-info/METADATA,sha256=aFNUX6uUZZHS7lPbYBTmMSH6TBiWmXbEgQNxPZNWiRI,17021
+pyrddlgym_jax-2.2.dist-info/WHEEL,sha256=jB7zZ3N9hIM9adW7qlTAyycLYW9npaWKLRzaoVcLKcM,91
+pyrddlgym_jax-2.2.dist-info/entry_points.txt,sha256=Q--z9QzqDBz1xjswPZ87PU-pib-WPXx44hUWAFoBGBA,59
+pyrddlgym_jax-2.2.dist-info/top_level.txt,sha256=n_oWkP_BoZK0VofvPKKmBZ3NPk86WFNvLhi1BktCbVQ,14
+pyrddlgym_jax-2.2.dist-info/RECORD,,

{pyRDDLGym_jax-2.0.dist-info → pyrddlgym_jax-2.2.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (75.8.0)
+Generator: setuptools (75.8.2)
 Root-Is-Purelib: true
 Tag: py3-none-any