undoing test changes #974
undoing test changes #974
Conversation
| .Append(new Ova(env, sdcaTrainer)) | ||
| .Append(new KeyToValueEstimator(env, "PredictedLabel")); | ||
|
|
||
| var model = pipeline.Fit(data); |
There was a problem hiding this comment.
Do we want to do something with the model to ensure it was made correctly?
There was a problem hiding this comment.
was thinking the same about all the tests we added recently. I'll so a sweep to all of them.
There was a problem hiding this comment.
The purpose of these 'tests' is generally to demonstrate how we perform the scenario listed in the summary. I think it's enough that the code compiles and runs without throwing.
There are separate tests for estimators, trainers etc., that are designed to verify that the trained models behave as expected. Making every test validate everything is, I think, not necessary.
In reply to: 219552007 [](ancestors = 219552007)
There was a problem hiding this comment.
My thoughts:
Do those separate tests do the exact same thing as these? If yes, then these tests are redundant. If no, then these tests should be validating the scenario works as expected because the separate tests are doing something different.
There was a problem hiding this comment.
I will compact them. Thanks @eerhardt.
|
|
||
| var testLoader = TextLoader.ReadFile(env, MakeIrisTextLoaderArgs(), new MultiFileSource(dataPath)); | ||
| var testData = testLoader.AsEnumerable<IrisDataNoLabel>(env, false); | ||
| var testData = testLoader.AsEnumerable<IrisData>(env, false); |
There was a problem hiding this comment.
IrisData [](start = 55, length = 8)
@Zruty0 instead?
|
This PR does not reference an issue. Please fix this before merging. #Resolved |
ghost
locked as resolved and limited conversation to collaborators
Fixing merge errors on those two files, on the TreeEstimators PR.