A simple library that, among other things, provides a simple on-disk cache is joblib. The package can be used in a similar way as lru_cache, except that the results will be stored on disk and will persist between runs.

The joblib module provides the Memory class that can be used to memoize functions using the Memory.cache decorator:

    from joblib import Memory
    memory = Memory(cachedir='/path/to/cachedir')

    @memory.cache
    def sum2(a, b):
        return a + b

The function will behave similar to lru_cache, with the exception that the results will be stored on-disk in the directory specified by the cachedir argument during Memory initialization. Additionally, the cached results will persist over subsequent runs!

The Memory.cache method also allows to limit recomputation only when certain arguments change, and the resulting decorated function supports basic functionalities to clear and analyze the cache. 

Perhaps the best joblib feature is that, thanks to intelligent hashing algorithms, it provides efficient memoization of functions that operate on numpy arrays, and is particularly useful in scientific and engineering applications.