Li Zhang

Results 72 comments of Li Zhang
trafficstars

<Interval inference> slower than <continuous inference> when use faster-rcnn TRT model C++ mmdeploy inference server!!!

This is why there are warm-up iterations in speed benchmarks. The GPU frequency is lower when it's in an idle state. You may try to lock the GPU performance level...

<Interval inference> slower than <continuous inference> when use faster-rcnn TRT model C++ mmdeploy inference server!!!

In my test results, inserting a 2 second interval incurs 10-30% more latency.

Error when build demo on linux: /usr/bin/ld: /usr/lib/gcc/x86_64-linux-gnu/7/libstdc++fs.a(ops.o): undefined reference to symbol '_ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE6resizeEmc'

Hi, What's your Ubuntu version and how did you install GCC-7?

Error when build demo on linux: /usr/bin/ld: /usr/lib/gcc/x86_64-linux-gnu/7/libstdc++fs.a(ops.o): undefined reference to symbol '_ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE6resizeEmc'

I followed the instructions in the doc using a fresh Ubuntu 20.04 environment and can't reproduce the problem. The symbol `_ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE6resizeEmc` demangles to `std::__cxx11::basic_string::resize(unsigned long, char) ` which is from...

Could you please support MMTracking for TensorRT deployment?

Deploying SOT algorithms from mmtracking may not be easy. As you can see, mmtracking split them into 2 functions: `init` and `track`. Although they share some common logic, many steps...

Confused when reading the source code

Sorry for the confusion. The API for inference backend and post-processing is not part of the public API and is evolving in a fast pace.

how to do batch inference?

Hi, can you check if #839 solves your problem?

[Enhancement] Refactor SDK pipeline

> * How to set the batch size? `NetModule` has nothing to do with batch size for now. > * If one detection task is followed by two classification tasks,...

Tensorrt batch inference is much slower when batch size is larger.

You may find this helpful https://github.com/open-mmlab/mmdeploy/issues/839#issuecomment-1206029364

Tensorrt batch inference is much slower when batch size is larger.

The segmentation model is large enough to saturate your device, don't expect large speed up using batch inference.