Use AWS to conduct a massive search for embeddings for a hard problem?
I'm using minorminer.find_embedding() to search for embeddings for increasingly hard problems. I'm looking at max chain length and average chain length as suggested by Alex. It took my MacBook Pro almost six hours to try 100 times to find an embedding for a particular problem. Three embeddings were found, with an embedding length of 424, and in the best case a max chain length of 14 and an average chain length of 3.337. QPU time is scarce and expensive, but CPU time is abundant and cheap. I'd like to improve my chances of getting good results from the QPU by finding many embeddings and using the few with the most promising metrics. This leads me to think of using Amazon Web Services to conduct a massive search for embeddings in a reasonable wall-clock time.
Does anyone have any comments about this idea?