Describe the bug
It seems there is an issue related to how much metadata or context is passed when hashing a DictConfig object. As a consequence, hashes of two DictConfig objects differ, in cases where the objects are in fact equal. We have observed this in cases where the 2 objects are "sub configs" of the parent DictConfig object.
Describe the bug It seems there is an issue related to how much metadata or context is passed when hashing a DictConfig object. As a consequence, hashes of two DictConfig objects differ, in cases where the objects are in fact equal. We have observed this in cases where the 2 objects are "sub configs" of the parent DictConfig object.
To Reproduce
Expected behavior I would expect all 4 assert statements to pass. It is unexpected that
hash(d.a) != hash(d.b)
whend.a == d.b
.Additional context