Manual cache handling

[Richienb]

To best utilize the GPU, my inputs should be processed in batches (speedup ~20 hours to ~2 hours). So, I want to run the function on the inputs in batches and manually update the cache with the returned values, so that later processes that take them one by one would be pulling from the cache. - am requesting a public API for this.

[Richienb]

Short guide using internal methods in 1.3.2:

from joblib import Memory

mem = Memory('.cache')

f = mem.cache(expensive_operation)

ids = f._get_output_identifiers(*args, **kwargs)

# Check if in cache
f._is_in_cache_and_valid(ids)

# Store in cache
f.store_backend.dump_item(ids, f(*args, **kwargs), verbose=f._verbose)