Multivartiate Anomaly Detector - keeps giving InferenceLostConnection and Timeout errors during detect step

Question

Hi,

I am trying out the Python API for Multivariate Anomaly Detection. Model training succeeds without issues, but in the detect step I keep getting InferenceLostConnection and Timeout errors.

I have found that the InferenceLostConnection error can be avoided if I supply at most 3,000 data points per API call (despite the maximum number of data points per inference API call being 20,000). However, I work with relatively high frequency data (30 sec.) and this is not a feasible solution for me. The Timeout errors seem to occur at random and I have not been able to find a workaround. I've also opened a github issue for this here https://github.com/Azure-Samples/AnomalyDetector/issues/48.

Any idea what is causing this and if it will be fixed soon? I would love to try out this service properly to see what it can do, but right now it seems too unstable...

Best,
Cheryl