Hi @wang-xinyu , I tried to update the createEngine method and faced an issue. If you can help me resolve it would be wonderful.
So, this is the updated method according to the updated yolov5s model:
root@c5de50828b33:/home/build# ./yolov5s -s
Loading weights: ../yolov5s.wts
[06/26/2020-03:50:24] [E] [TRT] (Unnamed Layer* 132) [Deconvolution]: kernel weights has count 1024 but 262144 was expected
[06/26/2020-03:50:24] [E] [TRT] (Unnamed Layer* 132) [Deconvolution]: count of 1024 weights in kernel, but kernel dimensions (2,2) with 256 input channels, 256 output channels and 1 groups were specified. Expected Weights count is 256 * 2*2 * 256 / 1 = 262144
[06/26/2020-03:50:24] [E] [TRT] (Unnamed Layer* 132) [Deconvolution]: kernel weights has count 1024 but 262144 was expected
[06/26/2020-03:50:24] [E] [TRT] (Unnamed Layer* 132) [Deconvolution]: count of 1024 weights in kernel, but kernel dimensions (2,2) with 256 input channels, 256 output channels and 1 groups were specified. Expected Weights count is 256 * 2*2 * 256 / 1 = 262144
[06/26/2020-03:50:24] [E] [TRT] (Unnamed Layer* 132) [Deconvolution]: kernel weights has count 1024 but 262144 was expected
[06/26/2020-03:50:24] [E] [TRT] (Unnamed Layer* 132) [Deconvolution]: count of 1024 weights in kernel, but kernel dimensions (2,2) with 256 input channels, 256 output channels and 1 groups were specified. Expected Weights count is 256 * 2*2 * 256 / 1 = 262144
[06/26/2020-03:50:24] [E] [TRT] (Unnamed Layer* 132) [Deconvolution]: kernel weights has count 1024 but 262144 was expected
[06/26/2020-03:50:24] [E] [TRT] (Unnamed Layer* 132) [Deconvolution]: count of 1024 weights in kernel, but kernel dimensions (2,2) with 256 input channels, 256 output channels and 1 groups were specified. Expected Weights count is 256 * 2*2 * 256 / 1 = 262144
[06/26/2020-03:50:24] [E] [TRT] (Unnamed Layer* 132) [Deconvolution]: kernel weights has count 1024 but 262144 was expected
[06/26/2020-03:50:24] [E] [TRT] (Unnamed Layer* 132) [Deconvolution]: count of 1024 weights in kernel, but kernel dimensions (2,2) with 256 input channels, 256 output channels and 1 groups were specified. Expected Weights count is 256 * 2*2 * 256 / 1 = 262144
[06/26/2020-03:50:24] [E] [TRT] (Unnamed Layer* 132) [Deconvolution]: kernel weights has count 1024 but 262144 was expected
[06/26/2020-03:50:24] [E] [TRT] (Unnamed Layer* 132) [Deconvolution]: count of 1024 weights in kernel, but kernel dimensions (2,2) with 256 input channels, 256 output channels and 1 groups were specified. Expected Weights count is 256 * 2*2 * 256 / 1 = 262144
[06/26/2020-03:50:24] [E] [TRT] (Unnamed Layer* 132) [Deconvolution]: kernel weights has count 1024 but 262144 was expected
[06/26/2020-03:50:24] [E] [TRT] (Unnamed Layer* 132) [Deconvolution]: count of 1024 weights in kernel, but kernel dimensions (2,2) with 256 input channels, 256 output channels and 1 groups were specified. Expected Weights count is 256 * 2*2 * 256 / 1 = 262144
[06/26/2020-03:50:24] [E] [TRT] (Unnamed Layer* 132) [Deconvolution]: kernel weights has count 1024 but 262144 was expected
[06/26/2020-03:50:24] [E] [TRT] (Unnamed Layer* 132) [Deconvolution]: count of 1024 weights in kernel, but kernel dimensions (2,2) with 256 input channels, 256 output channels and 1 groups were specified. Expected Weights count is 256 * 2*2 * 256 / 1 = 262144
Building engine, please wait for a while...
[06/26/2020-03:50:24] [E] [TRT] (Unnamed Layer* 132) [Deconvolution]: kernel weights has count 1024 but 262144 was expected
[06/26/2020-03:50:24] [E] [TRT] (Unnamed Layer* 132) [Deconvolution]: count of 1024 weights in kernel, but kernel dimensions (2,2) with 256 input channels, 256 output channels and 1 groups were specified. Expected Weights count is 256 * 2*2 * 256 / 1 = 262144
[06/26/2020-03:50:24] [E] [TRT] Could not compute dimensions for (Unnamed Layer* 132) [Deconvolution]_output, because the network is not valid
[06/26/2020-03:50:24] [E] [TRT] Network validation failed.
Build engine successfully!
yolov5s: /home/yolov5s.cpp:233: void APIToModel(unsigned int, nvinfer1::IHostMemory**): Assertion `engine != nullptr' failed.
Aborted (core dumped)
I am facing problems in the updated Deconvolution which now takes input from a ConvBnLeaky block rather than bottleneck CSP
Any help would be great. Thanks!
Hi @wang-xinyu , I tried to update the createEngine method and faced an issue. If you can help me resolve it would be wonderful. So, this is the updated method according to the updated yolov5s model:
And when I run ./yolov5s -s I get this:
I am facing problems in the updated Deconvolution which now takes input from a ConvBnLeaky block rather than bottleneck CSP Any help would be great. Thanks!