Welcome to the official documentation for JAxtar. This documentation provides detailed information about the available commands and their options.
JAxtar is operated through a command-line interface. Below is a list of the main commands available.
These commands are used to solve puzzles using different search algorithms.
astar: Solves a puzzle using the A* search algorithm.astar_d: Solves a puzzle using the A* Deferred search algorithm.bi_astar: Solves a puzzle using the Bidirectional A* search algorithm.bi_astar_d: Solves a puzzle using the Bidirectional A* Deferred search algorithm.id_astar: Solves a puzzle using the Iterative Deepening A* search algorithm.beam: Solves a puzzle using Beam Search.qstar: Solves a puzzle using the Q* search algorithm, guided by a Q-function.bi_qstar: Solves a puzzle using the Bidirectional Q* search algorithm.id_qstar: Solves a puzzle using the Iterative Deepening Q* search algorithm.qbeam: Solves a puzzle using the Q-Beam search algorithm.
human_play: Allows you to play a puzzle interactively in the terminal.
These commands are used to train neural network models for heuristic search.
distance_train heuristic: Trains a neural network to act as a heuristic function by predicting the distance to the goal.distance_train qfunction: Trains a neural network to serve as a Q-function for estimating action costs.world_model_train train: Trains a discrete world model that learns the puzzle's transition dynamics.world_model_train make_transition_dataset: Generates a dataset of state transitions for training world models.world_model_train make_sample_data: Generates a dataset of initial and target state pairs.world_model_train make_eval_trajectory: Generates a long trajectory for evaluating world model predictions.
These commands are used to evaluate trained models and search algorithms.
eval heuristic: Evaluates heuristics usingastar,astar_d, orbeamsearch.eval qlearning: Evaluates Q-functions usingqstarorqbeamsearch.benchmark: Evaluates search strategies on standardized benchmark datasets.benchmark logging: Explanation of logs, output files, and metrics generated by evaluations.benchmark results: A record of experimental results.