[tune] implement shim instantiation

[sumanthratna]

Why are these changes needed?

makes creating search algorithms based on a string easier (useful for CLIs where users define search alg and scheduler to use)

Related issue number

related: #9969, #10401, #10444, #10451 (closes #10451)

Checks

• I've run scripts/format.sh to lint the changes in this PR.
• I've included any doc changes needed for https://docs.ray.io/en/latest/.
• I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failure rates at https://ray-travis-tracker.herokuapp.com/.
• Testing Strategy
  • Unit tests
  • Release tests
  • This PR is not tested (please justify below)

CC @richardliaw

[sumanthratna]

yikes this PR's commits have become way more bloated than they need to be; @richardliaw should I rebase and force-push to reduce the number of commits?

[richardliaw]

Hmm I don't really mind - up to you!

[richardliaw]

Looks great! Do you mind adding this into the docs too? ray/docs/source/tune/api_docs/{suggestion / schedulers}.rst

[richardliaw]

bump - looks like tests are failing. I think you should do type(scheduler) == type(...) in the unit tests

[sumanthratna]

Looks great! Do you mind adding this into the docs too? ray/docs/source/tune/api_docs/{suggestion / schedulers}.rst

sure, I'll try to do that soon

looks like tests are failing. I think you should do type(scheduler) == type(...) in the unit tests

Oh—I had (lazily) assumed that schedulers overrode the equality operator. Is this something we should implement? I don't think it would be too hard (untested):

def __eq__(self, other):
    if type(self) != type(other):
        return False
    return self.__dict__ == other.__dict__

this is probably better suited for a different PR, though

[richardliaw]

Hm I guess I don't know if those semantics really are needed since you're almost certainly only using 1 scheduler.

[richardliaw]

Thanks for putting together this PR though! Definitely looking forward to getting this merged :)

[sumanthratna]

@richardliaw can you take a look at 6dcf1f0 and let me know if I should change anything else? once you take a look I'll add to the schedulers doc too

[richardliaw]

looks good!

…

On Fri, Sep 4, 2020 at 6:05 AM Sumanth Ratna ***@***.***> wrote: @richardliaw <https://github.com/richardliaw> can you take a look at 6dcf1f0 <6dcf1f0> and let me know if I should change anything else? once you take a look I'll add to the schedulers doc too — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#10456 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABCRZZID5ZPPR6XURMLA7B3SEDQYPANCNFSM4QRAOXYA> .

[richardliaw]

Hey @sumanthratna looks like the create search alg test is still broken

Don't forget to run ci/format.sh too to address lint test!

[sumanthratna]

Hey @sumanthratna looks like the create search alg test is still broken

@richardliaw ugh I had copied the example from the docs because I wanted to write the test correctly but I was looking at the 0.8.7 docs instead of master...fix is (hopefully) on the way

[richardliaw]

hmm tests still failing?

…

On Fri, Sep 4, 2020 at 4:04 PM Sumanth Ratna ***@***.***> wrote: Hey @sumanthratna <https://github.com/sumanthratna> looks like the create search alg test is still broken @richardliaw <https://github.com/richardliaw> ugh I had copied the example from the docs because I wanted to write the test correctly but I was looking at the 0.8.7 docs instead of master...fix is (hopefully) on the way — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#10456 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABCRZZJMYNRHETM64T2WLZLSEFW7RANCNFSM4QRAOXYA> .

[sumanthratna]

we can wait until all tests finish but @richardliaw I think this is finally ready for merging