On "all 2^39 patterns" — any explicit file is infeasible (minimum ~2.75 TB). The correct answer is: the pattern space is [0, 2^39). A stream_all_patterns() generator is included that yields every pattern on-demand with zero memory overhead — decode pattern k directly from integer k without storing anything.
In short: use TopoDevPOC_n39.py if you dont want unnessesary file saved on your disk. And use TopoDevPOC_n39_save.py if you want to store every pattern on to CSV format with a total of 2.75… See the full description on the dataset page:
https://huggingface.co/datasets/algorembrant/549billion_directional-TopoDevPOC-patterns.
