What you would like to be added?
Instead of having individual tests to verify build objects, we might want to run EnforceMLPolicy() and Build() functions in TestFlux test akin to how MPI does it. See
|
err = p.(framework.EnforceMLPolicyPlugin).EnforceMLPolicy(tc.info, tc.trainJob) |
|
if diff := gocmp.Diff(tc.wantMLPolicyError, err, cmpopts.EquateErrors()); len(diff) != 0 { |
|
t.Errorf("Unexpected error from EnforceMLPolicy (-want, +got): %s", diff) |
|
} |
|
if diff := gocmp.Diff(tc.wantInfo, tc.info, |
|
cmpopts.SortSlices(func(a, b string) bool { return a < b }), |
|
cmpopts.SortMaps(func(a, b int) bool { return a < b }), |
|
utiltesting.PodSetEndpointsCmpOpts, |
|
); len(diff) != 0 { |
|
t.Errorf("Unexpected info from EnforceMLPolicy (-want, +got): %s", diff) |
|
} |
|
var objs []apiruntime.ApplyConfiguration |
|
objs, err = p.(framework.ComponentBuilderPlugin).Build(ctx, tc.info, tc.trainJob) |
|
if diff := gocmp.Diff(tc.wantBuildError, err, cmpopts.EquateErrors()); len(diff) != 0 { |
and discussion
#3064 (comment)
Why is this needed?
It mirrors the testing strategy of the MPI plugin.
Love this feature?
Give it a 👍 We prioritize the features with most 👍
