Closed bvdmitri closed 1 year ago
@albertpod ping
Patch coverage: 91.30
% and project coverage change: +0.40
:tada:
Comparison is base (
8a89b44
) 79.88% compared to head (94babb1
) 80.28%.
:mega: This organization is not using Codecov’s GitHub App Integration. We recommend you install it so Codecov can continue to function properly for your repositories. Learn more
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Do you have feedback about the report comment? Let us know in this issue.
The before/after_iteration
existed before and their purpose is not necessarily to halt the iterations. Previously, they only allowed to do some extra actions (e.g. logging execution), but now they also allow to halt.
This PR fixes #84. @John-Boik
After this PR it is possible to return a
true/false
value from thebefore_iteration/after_iteration
callbacks.true
value indicates that the inference procedure should be halted and no further inference should be performed.The solution here is not ideal. Ideally we should support different "convergence" strategies, e.g based on BFE values. I like the proposal from @John-Boik in the #84, but I believe it must be a separate issue. For now I believe its fine to use callbacks that allow to halt iterations based on
true/false
.