(Question) How to use MetaQNL with multiple CPUs (Parallel processing)?

princeton-vl / MetaQNL

Learning Symbolic Rules for Reasoning in Quasi-Natural Language: https://arxiv.org/abs/2111.12038

BSD 2-Clause "Simplified" License

6 stars 0 forks source link

(Question) How to use MetaQNL with multiple CPUs (Parallel processing)? #21

Open realharryhero opened 1 month ago

realharryhero commented 1 month ago

Hello (If you have time!),

The tool is quite clean (no warnings on the large SCAN dataset)! When I ran it, though, checking top, it only ran on one CPU. In the paper, there was some training MetaQNL on four CPUs; I also checked train.jl --help but parallel computing did not appear. Is there a way to run MetaQNL on multiple CPUs? Thank you!

realharryhero commented 3 weeks ago

(bump!)

yangky11 commented 3 weeks ago

Sorry for the late response.

The machine we used for running the experiments has four CPUs, though the codebase doesn't effectively use multiple CPUs. The most time-consuming part in training is calling the MaxSAT solver (Z3 by default). It may be possible to run Z3 on multiple CPUs, though we didn't experiment with it.