Remove public 'fname' learner attribute

[basnijholt]

(original issue on GitLab)

opened by Joseph Weston (@jbweston) at 2018-11-15T16:32:45.712Z

The docs say that there are 2 ways to specify the name to save a learner's data: https://adaptive.readthedocs.io/en/latest/tutorial/tutorial.advanced-topics.html

I am not sure why fname is stored on the learner, rather it should only be a parameter to save and load and nothing else

[basnijholt]

originally posted by Bas Nijholt (@basnijholt) at 2018-11-19T13:32:18.497Z on GitLab

This is explained in https://gitlab.kwant-project.org/qt/adaptive/merge_requests/118.

[basnijholt]

originally posted by Joseph Weston (@jbweston) at 2018-11-19T13:45:33.747Z on GitLab

I still don't really see why it's necessary. BalancingLearner.load could take something additional that maps learners to filenames, for example. The API for 'BalancingLearner.load' is already not the same as for the other learners, because it expects a path to a directory rather than to a regular file, so it doesn't seem bad to have it also differ in other ways.

For example, for BalancingLearner, 'fname' could expect a callable that takes a learner and gives back a filename where the data can be got.

This means that your example from gitlab:!118:

def combo_fname(val):
    return '__'.join([f'{k}_{v}' for k, v in val.items()])

combos = adaptive.utils.named_product(a=[1, 2], b=[1])
learners = []
for combo in combos:
    l = Learner(partial(f, combo=combo))
    l.fname = combo_fname(combo)
    learners.append(l)
learner = BalancingLearner(learners)
learner.load(folder='data_folder')

becomes:

combos = adaptive.utils.named_product(a=[1, 2], b=[1])
learners = {Learner(partial(f, combo=combo)): combo for combo in combos}

def combo_fname(learner):
    val = learners[learner]
    return 'data_folder/' + '__'.join([f'{k}_{v}' for k, v in val.items()])

learner = BalancingLearner(list(learners.keys()))
learner.load(fname=combo_fname)

[basnijholt]

originally posted by Joseph Weston (@jbweston) at 2018-11-19T16:15:02.731Z on GitLab

In fact it can be even simpler because BalancingLearner.from_product uses functools:

def combo_fname(learner):  # this function is pure
    val = learner.function.keywords  # from functools.partial
    return 'data_folder/' + '__'.join([f'{k}_{v}' for k, v in val.items()])

combos = dict(a=[1, 2], b=[1])
learner = BalancingLearner.from_product(f, Learner1D, dict(bounds=(-1, 1)), combos)
learner.load(fname=combo_fname)