So when I try to run the following code

import alphacube

alphacube.load(model_id="small") # the default model

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)

the output would give me these errors

Traceback (most recent call last): File "/Users/vincent/Rubik Cube Solver/main.py", line 4, in <module> alphacube.load(model_id="small") # the default model ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages/alphacube/__init__.py", line 58, in load _solver.load(*args, **kwargs) File "/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages/alphacube/solver.py", line 55, in load self.model = load_model(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages/alphacube/model.py", line 55, in load_model state_dict = torch.load(model_path, map_location=torch.device("cpu")) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages/torch/serialization.py", line 993, in load with _open_zipfile_reader(opened_file) as opened_zipfile: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages/torch/serialization.py", line 447, in __init__ super().__init__(torch._C.PyTorchFileReader(name_or_buffer)) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ RuntimeError: PytorchStreamReader failed reading zip archive: failed finding central directory

I did pip install alphacube already

Hi there!

It is me again...

First of all, thank you very much for listening to my ideas and being so patient and nice with me.

I'm attempting Symbolic Regression on your model. However, managing 54 entries and 18 outputs is quite overwhelming. I'm considering training another model to predict the number of moves required for the AlphaCube to solve a given scramble, possibly with a small beam width to expedite data collection (I remember you did something like this, but I'm unsure if you have the model saved). For every solution with a length N, I would generate N data points, capturing the state after the scramble and the number of moves required to solve from the start to the end of the solution. Then, I'll utilize it as an evaluation function, like chess engines, and proceed with symbolic regression on this model, which involves 54 entries and 1 output. My intuition suggests it might contain some information on solving a Rubik's Cube within its weights. Additionally, it would be interesting to observe from the Symbolic Regression whether redundancies emerge, such as knowing two colors of a corner piece being sufficient to determine the entire piece.

Thank you!

Hi there, I'm testing a way to implement a FMC technique to search others solutions based on a single one. It is called NISS. From the example solve you put in "usage" here are the outputs:

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

Original output: [21] D L D2 R' U2 D B' D' U2 B U2 B' U' B2 D B2 D' B2 F2 U2 F2

New output:
[20] U2 R L' U2 F2 L' U2 F2 D' L U R2 F2 D' R D' B' R B' R'
[21] D R2 D B' D' B L' B' R F L2 R F' D' R B D2 R L2 U' D

With a higher beam width:

scramble="D U F2 L2 U' B2 F2 D L2 U R' F' D R' F' U L D' F' D R2"
beam_width=11024

Original output: [20] R' F L2 D R2 U' B' L' U2 F2 U L D U2 B2 D2 R2 U2 B2 R2

New output:
[19] R' F R B2 F2 R2 D2 F2 D2 B2 R' D B R L U2 R' D U
[19] D2 R U' F U2 F' U F2 R2 D2 R2 L U L' U2 D' R' F2 R'
[18] D2 L F' D F2 D' F D2 R2 B2 R' U L' U2 D' R' F2 R'

I did it by applying NISS with gradually increasing sequence of the original solution, from the beginning to the end, and finding a new solution on the inverse scramble, like we usually do in FMC.