difference between SearchClickModelTask and FitClickModelTask?

[lubeidi]

From what I understand, these two tasks run together using the same dataset and ultimately calls the same function i.e. func (fm *FM) Fit(). The difference is that SearchClickModelTask will use random search parameters and takes O(fm.nEpochs)*O(numTrials) while FitClickModelTask takes O(fm.nEpochs).

Since both two put a readlock on the clickDataset, runLoadDatasetTask() won't continue until both are done. When dataset is huge and FitClickModelTask takes several hours, SearchClickModelTask will take even days to complete, O(numTrials) times longer to be exact.

So my question is

1. In situations above, to reduce the time waiting for the SearchClickModelTask. Is it OK to put different numTrials in parallel? Will future Gorse release update this feature?
2. If I set numTrials to 1, then essentially SearchClickModelTask and FitClickModelTask are the same and FitClickModelTask results only should be enough. Is that correct?

[lubeidi]

Just figure out there are several misunderstandings of the system, these two tasks will block each other, and SearchClickModelTask acts like a hyperparameter-automl.

Now I have only one question: To reduce the time waiting for the SearchClickModelTask, is it OK to put different numTrials in parallel? Will future Gorse release update this feature?