DeepStream 6.0.1 Triton GRPC memory leak

In your config, nvinfer and nvinferserver used different model.

For the sample application yes, the difference in leak amount I meant for the custom application I cannot share. I didn’t try using YOLOv4 directly with nvinfer for the sample application.

We have created internal bug to track this mem leak issue, thanks for reporting this issue.

Hello, now use the latest nvcr.io/nvidia/deepstream: version 6.1 triton mirror in nvinferserver with tritonserver Grpc data interactive way, there is a memory leak, any improvement or patch for this problem? thank you!