Video memory leak

[XL634663985]

Description

When I use "with TrtRunner(self.onnx_session) as runner" for inference, there is no video memory leak, but the inference speed is very slow (because it has to be reloaded every time). If the model is set as a global variable, and the model is activated using "runner.activate()" and remains unchanged, it will cause video memory leak. How to solve it?

Environment

TensorRT Version:10.2.0.19

NVIDIA GPU:3090

NVIDIA Driver Version:560.94

CUDA Version:cuda_12.1.0_531.14_windows

CUDNN Version:cudnn-windows-x86_64-9.2.1.18_cuda12-archive

Operating System:windows10

Python Version (if applicable):3.11.9

Tensorflow Version (if applicable):

PyTorch Version (if applicable):

Baremetal or Container (if so, version):

Relevant Files

Model link:

Steps To Reproduce

Commands or scripts:

Have you tried the latest release?:

Can this model run on other frameworks? For example run ONNX model with ONNXRuntime (polygraphy run <model.onnx> --onnxrt):

[poweiw]

Hello! We'll need a few more information to reproduce:

1. Link to your onnx model
2. A script to reproduce the error
3. The source you saw the vram memory leak, whether it's from cuda-memcheck, cuda-gdb