FLAML
FLAML copied to clipboard
arima and sarimax init value
trafficstars
In version 1.0.2 to 1.0.5, the init value of quniform HPs in arima and sarimax is int which doesn't pass the validity check. So these init values are not respected.
I propose a change in PR #575 .
Also, I suggest adding a performance test like https://github.com/microsoft/FLAML/blob/619107edf504905b069e0ca77f117f8aca32d042/test/automl/test_notebook_example.py#L73 for the forecasting task (@int-chaos ) as well as the nlp tasks (@liususan091219 ) to prevent similar issues to happen.
Hi I see a merged PR for this issue Is there something left to close this issue?