getBindingIndex() is DEPRECATED

[Egorundel]

Description

Hello!
I want to find out the BindingIndex by the tensor name.
How can I do this, provided that the getBindingIndex() method is deprecated in 8.5.

Environment

TensorRT Version: 8.6

NVIDIA GPU: RTX 3060

NVIDIA Driver Version: 545.23.06

CUDA Version: 11.1

CUDNN Version: 8.0.4

Operating System:

Python Version (if applicable): 3.8

PyTorch Version (if applicable): 10.1

[lix19937]

If you just want to get shape(dims) of tensor(binding), use getTensorShape(char const * tensorName), when TensorRT Version is 8.6

Ref https://docs.nvidia.com/deeplearning/tensorrt/api/c_api/classnvinfer1_1_1_i_cuda_engine.html#af96a2ee402ab47b7e0b7f0becb63d693

[DeepLearnerYe]

getTensorShape(char const * tensorName) requires tensorName. If I want to get the name automatically instead of specifying the name, getBindingName(int32_t) is provided. But it is also deprecated in 8.5, what can I do?

[lix19937]

@DeepLearnerYe
Use

int32_t nvinfer1::ICudaEngine::getNbIOTensors()	
char const * nvinfer1::ICudaEngine::getIOTensorName(int32_t index) const

ref in https://docs.nvidia.com/deeplearning/tensorrt/archives/tensorrt-853/api/index.html