server
server copied to clipboard
Show ensemble stage where error happens
For now, if an error happens at some stage of the ensemble it prints something like tritonclient.utils.InferenceServerException: [StatusCode.INTERNAL] in ensemble 'tts', PyTorch execute failure: The following operation failed in the TorchScript interpreter.
So, there is no chance to understand which model in the ensemble was at fault, so my proposal is to print the ensemble step, where the error happens.
Good find, Alexander. Thank you. I've filed a ticket to look into this feature request.