alphacube 0.1.1

A powerful & flexible Rubik's Cube solver

AlphaCube

AlphaCube is a powerful & flexible Rubik's Cube solver that extends EfficientCube.

App

We offer an interactive demo application for you to try out AlphaCube: alphacube.dev

Installation

Open a terminal and execute the following command:

pip install alphacube

Usage

Basic

Using AlphaCube in Python is as simple as this:

import alphacube

# Load a trained DNN
alphacube.load(model_id="small") # the default model

# Solve the cube using a given scramble
result = alphacube.solve(
    scramble="D U F2 L2 U' B2 F2 D L2 U R' F' D R' F' U L D' F' D R2",
    beam_width=1024,
)
print(result)

Output

{
    'solutions': [
        "D L D2 R' U2 D B' D' U2 B U2 B' U' B2 D B2 D' B2 F2 U2 F2"
    ],
    'num_nodes': 19744,
    'time':1.4068585219999659
}

AlphaCube selects the smallest model by default and takes about a few seconds to find moderately good solution(s).

Better Solutions

If you want even shorter solutions, simply increase the beam_width parameter:

alphacube.load() # model_id="small"
result = alphacube.solve(
    scramble="D U F2 L2 U' B2 F2 D L2 U R' F' D R' F' U L D' F' D R2",
    beam_width=65536,
)
print(result)

Output

{
    'solutions': [
        "D' R' D2 F' L2 F' U B F D L D' L B D2 R2 F2 R2 F'",
        "D2 L2 R' D' B D2 B' D B2 R2 U2 L' U L' D' U2 R' F2 R'"
    ],
    'num_nodes': 968984,
    'time': 45.690575091997744
}

You generally get more and better solutions, at the sacrifice of an increased wall-clock time and computational cost.

Applying Ergonomic Bias

ergonomic_bias = {
    "U": 0.9,   "U'": 0.9,  "U2": 0.8,
    "R": 0.8,   "R'": 0.8,  "R2": 0.75,
    "L": 0.55,  "L'": 0.4,  "L2": 0.3,
    "F": 0.7,   "F'": 0.6,  "F2": 0.6,
    "D": 0.3,   "D'": 0.3,  "D2": 0.2,
    "B": 0.05,  "B'": 0.05, "B2": 0.01,
    "u": 0.45,  "u'": 0.45, "u2": 0.4,
    "r": 0.3,   "r'": 0.3,  "r2": 0.25,
    "l": 0.2,   "l'": 0.2,  "l2": 0.15,
    "f": 0.35,  "f'": 0.3,  "f2": 0.25,
    "d": 0.15,  "d'": 0.15, "d2": 0.1,
    "b": 0.03,  "b'": 0.03, "b2": 0.01
} # Each value represents the desirability score

result = alphacube.solve(
    scramble="D U F2 L2 U' B2 F2 D L2 U R' F' D R' F' U L D' F' D R2",
    beam_width=65536,
    ergonomic_bias=ergonomic_bias
)
print(result)

Output

{
    'solutions': [
        "u' U' f' R2 U2 R' L' F' R D2 f2 R2 U2 R U L' U R L",
        "u' U' f' R2 U2 R' L' F' R D2 f2 R2 U2 R d F' U f F",
        "u' U' f' R2 U2 R' L' F' R u2 F2 R2 D2 R u f' l u U"
    ],
    'num_nodes': 1078054,
    'time': 56.13087955299852
}

GPU Acceleration

If you have a GPU, we highly recommend you select the largest model for the best possible performance:

alphacube.load(model_id="large")
result = alphacube.solve(
    scramble="D U F2 L2 U' B2 F2 D L2 U R' F' D R' F' U L D' F' D R2",
    beam_width=65536,
)
print(result)

Output

{
    'solutions': ["D F L' F' U2 B2 U F' L R2 B2 U D' F2 U2 R D'"],
    'num_nodes': 903448,
    'time': 20.46845487099995
}

You will get even better solutions in an order of magnitude smaller amount of time.

On the contrary, using a model larger than "small" (i.e., "base" or "large") on CPU will likely result in inferior temporal performance.

Verbose Mode

Additionally, you may call alphacube.set_verbose() to keep track of the progress. It will display the current depth of search on your terminal.

Please refer to our documentation for more, especially "Getting Started"

How It Works

At the heart of AlphaCube lies a methodology rooted in the paper titled "Self-Supervision is All You Need for Solving Rubik's Cube" (TMLR'23), the official code of which is also available as a GitHub repository. This project releases the three largest solvers trained in Half-Turn Metric (HTM) as described in Section 7. Scaling Law.

In EfficientCube, a Deep Neural Network (DNN) is trained to predict the probability distribution of the next moves that would bring a given state one step closer to the goal. These predicted moves are then applied sequentially to solve a given scramble.

This project releases the three largest, compute-optimally trained solvers as described in Section 7. Scaling Law.

You may also read "How It Works" in the documentation for illustrated descriptions.

Contributing

You are more than welcome to collaborate on AlphaCube. If you're interested, please read our CONTRIBUTING guide to get started.

License

AlphaCube is open-sourced under the MIT license, granting you the freedom to use and modify it.