[Q] How to call predictor backend independently from trasformer

[cceyda]

I'm following this docs example on using triton as the predictor with a transformer.
Is it possible to call the predictor (in this case triton server) directly when deployed together with a transformer?
Use case: I wanna use the transformer for pre-post processing sometimes. But sometimes I want to call the triton server directly

I tried a couple of things but I just can't figure out the right call address/port api signature. It always seems to route to the transformer.

[sivanantha321]

@cceyda Transformer and Predictor both have separate URLs. You can bypass the transformer by using the predictor URL. You can find it in the isvc status.