issues about the size of yolov5s-int8 and yolov5s-fp16

[JackjackFan]

Env

• GPU, e.g. V100, RTX2080, TX2, Xavier NX, Nano, etc.
• OS, e.g. Ubuntu16.04, Win10, etc.
• Cuda version
• TensorRT version

About this repo

• which branch/tag/commit are you using?
• which model? yolov5, retinaface?

Your problem

Hi,Mr.Wang. I try to convert yolov5s.wts to engie and quali it to int8.But besides the map of the difference between int8 and fp16.The size of model is different too. yolov5s-fp16 14.2M,and yolov5s-in8 14.3M.What is the reason of this difference?In my opinin,IN8 should be half of fp16 right?could you pleace give me a hand?Thanks a lot

• what is your command? e.g. sudo ./yolov5 -s
• what's your output?
• what output do you expect?

[rajeshroy402]

Hi, @JackjackFan - how did you convert? I have a custom model Yolov5 5.0 and I am getting errors while running the code on my Jetson Xavier NX env

~/tensorrtx/yolov5/build$ sudo ./yolov5 -s yolov5.wts yolov5.engine s
Loading weights: yolov5.wts
[04/15/2022-00:10:18] [E] [TRT] (Unnamed Layer* 201) [Convolution]: kernel weights has count 2304 but 32640 was expected
[04/15/2022-00:10:18] [E] [TRT] (Unnamed Layer* 201) [Convolution]: count of 2304 weights in kernel, but kernel dimensions (1,1) with 128 input channels, 255 output channels and 1 groups were specified. Expected Weights count is 128 * 1*1 * 255 / 1 = 32640
[04/15/2022-00:10:18] [E] [TRT] (Unnamed Layer* 201) [Convolution]: kernel weights has count 2304 but 32640 was expected
[04/15/2022-00:10:18] [E] [TRT] (Unnamed Layer* 201) [Convolution]: count of 2304 weights in kernel, but kernel dimensions (1,1) with 128 input channels, 255 output channels and 1 groups were specified. Expected Weights count is 128 * 1*1 * 255 / 1 = 32640
[04/15/2022-00:10:18] [E] [TRT] (Unnamed Layer* 201) [Convolution]: kernel weights has count 2304 but 32640 was expected
[04/15/2022-00:10:18] [E] [TRT] (Unnamed Layer* 201) [Convolution]: count of 2304 weights in kernel, but kernel dimensions (1,1) with 128 input channels, 255 output channels and 1 groups were specified. Expected Weights count is 128 * 1*1 * 255 / 1 = 32640
[04/15/2022-00:10:18] [E] [TRT] (Unnamed Layer* 201) [Convolution]: kernel weights has count 2304 but 32640 was expected
[04/15/2022-00:10:18] [E] [TRT] (Unnamed Layer* 201) [Convolution]: count of 2304 weights in kernel, but kernel dimensions (1,1) with 128 input channels, 255 output channels and 1 groups were specified. Expected Weights count is 128 * 1*1 * 255 / 1 = 32640
Building engine, please wait for a while...
[04/15/2022-00:10:18] [E] [TRT] (Unnamed Layer* 201) [Convolution]: kernel weights has count 2304 but 32640 was expected
[04/15/2022-00:10:18] [E] [TRT] (Unnamed Layer* 201) [Convolution]: count of 2304 weights in kernel, but kernel dimensions (1,1) with 128 input channels, 255 output channels and 1 groups were specified. Expected Weights count is 128 * 1*1 * 255 / 1 = 32640
[04/15/2022-00:10:18] [E] [TRT] Could not compute dimensions for (Unnamed Layer* 201) [Convolution]_output, because the network is not valid.
[04/15/2022-00:10:18] [E] [TRT] Network validation failed.
Build engine successfully!
yolov5: /home/nvidia/tensorrtx/yolov5/yolov5.cpp:243: void APIToModel(unsigned int, nvinfer1::IHostMemory**, bool&, float&, float&, std::__cxx11::string&): Assertion `engine != nullptr' failed.
Aborted

I will much appreciate if you can help.

[suyashhchougule]

Hi, @JackjackFan - how did you convert? I have a custom model Yolov5 5.0 and I am getting errors while running the code on my Jetson Xavier NX env

~/tensorrtx/yolov5/build$ sudo ./yolov5 -s yolov5.wts yolov5.engine s
Loading weights: yolov5.wts
[04/15/2022-00:10:18] [E] [TRT] (Unnamed Layer* 201) [Convolution]: kernel weights has count 2304 but 32640 was expected
[04/15/2022-00:10:18] [E] [TRT] (Unnamed Layer* 201) [Convolution]: count of 2304 weights in kernel, but kernel dimensions (1,1) with 128 input channels, 255 output channels and 1 groups were specified. Expected Weights count is 128 * 1*1 * 255 / 1 = 32640
[04/15/2022-00:10:18] [E] [TRT] (Unnamed Layer* 201) [Convolution]: kernel weights has count 2304 but 32640 was expected
[04/15/2022-00:10:18] [E] [TRT] (Unnamed Layer* 201) [Convolution]: count of 2304 weights in kernel, but kernel dimensions (1,1) with 128 input channels, 255 output channels and 1 groups were specified. Expected Weights count is 128 * 1*1 * 255 / 1 = 32640
[04/15/2022-00:10:18] [E] [TRT] (Unnamed Layer* 201) [Convolution]: kernel weights has count 2304 but 32640 was expected
[04/15/2022-00:10:18] [E] [TRT] (Unnamed Layer* 201) [Convolution]: count of 2304 weights in kernel, but kernel dimensions (1,1) with 128 input channels, 255 output channels and 1 groups were specified. Expected Weights count is 128 * 1*1 * 255 / 1 = 32640
[04/15/2022-00:10:18] [E] [TRT] (Unnamed Layer* 201) [Convolution]: kernel weights has count 2304 but 32640 was expected
[04/15/2022-00:10:18] [E] [TRT] (Unnamed Layer* 201) [Convolution]: count of 2304 weights in kernel, but kernel dimensions (1,1) with 128 input channels, 255 output channels and 1 groups were specified. Expected Weights count is 128 * 1*1 * 255 / 1 = 32640
Building engine, please wait for a while...
[04/15/2022-00:10:18] [E] [TRT] (Unnamed Layer* 201) [Convolution]: kernel weights has count 2304 but 32640 was expected
[04/15/2022-00:10:18] [E] [TRT] (Unnamed Layer* 201) [Convolution]: count of 2304 weights in kernel, but kernel dimensions (1,1) with 128 input channels, 255 output channels and 1 groups were specified. Expected Weights count is 128 * 1*1 * 255 / 1 = 32640
[04/15/2022-00:10:18] [E] [TRT] Could not compute dimensions for (Unnamed Layer* 201) [Convolution]_output, because the network is not valid.
[04/15/2022-00:10:18] [E] [TRT] Network validation failed.
Build engine successfully!
yolov5: /home/nvidia/tensorrtx/yolov5/yolov5.cpp:243: void APIToModel(unsigned int, nvinfer1::IHostMemory**, bool&, float&, float&, std::__cxx11::string&): Assertion `engine != nullptr' failed.
Aborted

I will much appreciate if you can help.

@rajeshroy402 I am having same issue , were you able to solve it ?

[sicarioakki]

I am facing the same issue, @JackjackFan @rajeshroy402 @suyashhchougule were you able to solve them?

[wang-xinyu]

Errors like this is due to model mismatch between pytorch and tensorrt. You can follow the readme to use the matched version, and try official model first.

[04/15/2022-00:10:18] [E] [TRT] (Unnamed Layer* 201) [Convolution]: kernel weights has count 2304 but 32640 was expected

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.