Aaqib

@hungtooc For returning custom data types, you can serialize it to a bytearray and then deserialize it on the client side. For instance, for numpy array, you could use `np.save`...

Hey @pankajvshrma Right now we only support [3 inference metrics](https://github.com/pytorch/serve/blob/b87d7e59790e06c74f486c186f4a1d47a557b963/frontend/server/src/main/java/org/pytorch/serve/metrics/format/prometheous/PrometheusMetricManager.java#L10-L12) on the prometheus metrics endpoint. Unless you run an inference the metrics will be empty (with a 200OK response). If...

@pankajvshrma Seems like a bug with workflow. Will look into this

@harshbafna Looking at the prometheus java client [code](https://github.com/prometheus/client_java/blob/b61dd232a504e20dad404a2bf3e2c0b8661c212a/simpleclient/src/main/java/io/prometheus/client/SimpleCollector.java#L106), it seems a metric is initialized during register only if doesn't have a label