-
Notifications
You must be signed in to change notification settings - Fork 169
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
BasicStatistics API change #1644
base: main
Are you sure you want to change the base?
Conversation
|
||
tol = fp32tol if res.dtype == np.float32 else fp64tol | ||
assert_allclose(gtr, res, rtol=tol) | ||
|
||
@pytest.mark.parametrize("queue", get_queues()) | ||
@pytest.mark.parametrize("dtype", [np.float32, np.float64]) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
What is the reason behind this reordering?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
what do you mean by reordering? i have just added a new test test_multiple_options_uniform to make sure it works well if result_options is a list type
8b6fe24
to
bcdf148
Compare
/intelci: run |
I'm wondering if BasicStatistics and IncrementalBasicStatistics shouldn't be two functions, possibly located in 'metrics' which returns a dict. @KulikovNikita @olegkkruglov @samir-nasibli |
d78b82a
to
8162e95
Compare
/intelci: run |
1 similar comment
/intelci: run |
b33e801
to
b0c7ad2
Compare
/intelci: run |
This would drop basic statistics compute completely and break API - I guess we would need to deprecate it first? |
I would say that it is not breaking but more like fixing and making more consistent. As far as I understand, we have to use fit instead of compute in all estimators and compute was used here accidentally. But if we still need to use deprecation in such case then do we have any guidelines about how should it be done? |
Agreed that it is fixing, but that doesn't mean its not breaking - anyone that is using our BasicStatistics API will be impacted by this change. @Alexsandruss how do we generally handle this? (ie should we add deprecation warning message but leave compute() and just call fit() from it initially or can we remove compute() immediately) I would also expect that this may require an accompanying docs update, spmd interfaces, etc. |
We can leave |
I think this is good idea to discuss. In case if we will accept this, deprecation warnings also needed here. Personally I am good with current solution as well. |
@ethanglaser deprecation is added, spmd part is updated, but I can't see any docs containing information about old interfaces which should be updated according to these changes. Am I missing something? |
68ee924
to
d43a50d
Compare
/intelci: run |
@olegkkruglov please rebase you branch |
d43a50d
to
154ef88
Compare
/intelci: run |
Changes proposed in this pull request: