Error while converting VITS model to TensorRT engine

[Coastchb]

I use TTS to train a VITS model, and export the ONNX model(named coqui_vits.onnx).
And then use trtexec tool to convert onnx model to engine:
trtexec --onnx=coqui_vits.onnx --saveEngine=coqui_vits_bs1_sl128_fp16.engine --minShapes=input:1x128 --maxShapes=input:1x128 --optShapes=input:1x128 --fp16 --useCudaGraph

and got:
IExecutionContext::enqueueV3: Error Code 1: Cuda Runtime (operation not permitted when stream is capturing)

Could you please help me?