cqm.to_hash()

[hsadeghidw]

Application Often, we need to store the fingerprint of a CQM object. This may be useful when one needs to store a CQM object in memory or on disk or retrieve the result of previously solved CQM problems. Having a hash value that is sensitive to the values of all biases and constraints will make it very convenient to give a unique label to CQMs. I don't know if it's possible/useful to decide if it's also sensitive to labels, but I will personally always choose to be sensitive to labels

Proposed Solution

import hashlib

def _to_hash(cqm):
    with cqm.to_file() as f:
        return hashlib.sha1(f.read()).hexdigest()

Alternatives Required The method above is expensive and creates copies of the CQM object.

Additional Context The main use case to consider here is when a large problem is broken into smaller problems and many problems are submitted in parallel. In case any of them fail, a restart would ideally load previously solved problems and resubmit the failed ones. This may actually end up being a feature on its own. MultiCQMSampler?

[boothby]

There's a really cool graph hashing algorithm in networkx that might save significant effort here:

https://networkx.org/documentation/stable/reference/algorithms/generated/networkx.algorithms.graph_hashing.weisfeiler_lehman_graph_hash.html#networkx.algorithms.graph_hashing.weisfeiler_lehman_graph_hash