trt-fairmot
trt-fairmot copied to clipboard
liu-mengyang
About the speed result
thank you for the awesome contribution firstly. I implement this project successfully, and got the result below:
but when I test the inferenct time without using this project, i got the results below
| model | gpu_memroy_cost | ave_time | fps | model_inference | post_inference |
|---|---|---|---|---|---|
| farimot_dla34 | 1295MiB | 0.069 | 14 | 0.05 | 0.008 |
| farimot_dla34(MMCV) | 1591MiB | 0.062 | 14.57 | 0.038 | 0.008 |
As we can easily find that the acceleration is not much as we expected, and the gpu momery cost more than before.
Futhermore, I noticed when I run python compare_onnx_fairmot.py, the gpu memory hit 6700MB.
I think it need to check more carefully. I will be appreciate if anyone who can explain the results.
@zengjie617789 Sorry to bother you, it seems that you have managed to achieve the final fairmot.trt successfully. I have some troubles like this [E] Could not load plugin library: ./build/DCNv2PluginDyn.so, due to: libtorch.so: cannot open shared object file: No such file or directory when i run sh build_trt.sh to build trt model. But I do have the DCNv2PluginDyn.so in build directory and libtorch.so in torch's lib after compiling the plugins in the first step. Could you let me know how can you success to have that result please. Thank you in advance!
@PonyMaY I suggust you should post the version of TensorRT installed from source , as well as the pytorch. Futhermore, the details of DCNv2 installation; and have you export the ONNX model successfully? BTW, I didnot use these code to realize the fairmot-trt exactly. I used the DCN module from MMCV. The result I post here is for discussing the speed-up results.
@zengjie617789 Thank you for your prompt reply to me! Sorry to bother you again.
I did success to achieve the ONNX model(file sIze is 75MB, is it correct?) And the version of TensorRT I use is 7.2.2.3 and PyTorch 1.8.1+cu10.2. Someone in Issue #1 also use CUDA10.2 to run this repo.
Maybe i must use tensorrt:21.02-py3 of NGC TensorRT Docker to implement this project? So weird this error is and i can't find any hint or source of this error......