search

enazoe / yolo-tensorrt

TensorRT8.Support Yolov5n,s,m,l,x .darknet -> tensorrt. Yolov4 Yolov3 use raw darknet *.weights and *.cfg fils. If the wrapper is useful to you,please Star it.
MIT License
1.18k stars 316 forks source link

trt8 is not supply leaky? #187

Open tuteming opened 1 year ago

tuteming commented 1 year ago

trt8 is not supply leaky?

tuteming commented 1 year ago

I have a yolo-like cfg with leaky, it run in your trt7 version is good. but run in your trt8 version has wrong. error in chunk.cu line 74: NV_CUDA_CHECK(cudaMemcpyAsync((char)outputs[0] + b _n_size_split, (char)inputs[0] + b 2 * _n_size_split, _n_size_split, cudaMemcpyDeviceToDevice, stream));

may change leaky to mish in my config file?

tuteming commented 1 year ago

if in chunk.h set int _n_size_split=0
it can run, but no any detect results . you have any idea can help me?

thanks

tuteming commented 1 year ago

both in trt7(ok) and trt8(nothing result), in yolo.cpp, I use outputTensor.gridSize = (m_InputH / 32) pow(2, _n_yolo_ind); outputTensor.grid_h = (m_InputH / 32) pow(2, _n_yolo_ind); outputTensor.grid_w = (m_InputW / 32) * pow(2, _n_yolo_ind);

and delet if (m_NetworkType == "yolov4")//pan { outputTensor.gridSize = (m_InputH / 32) pow(2, 2 - _n_yolo_ind); outputTensor.grid_h = (m_InputH / 32) pow(2, 2 - _n_yolo_ind); outputTensor.grid_w = (m_InputW / 32) * pow(2, 2 - _n_yolo_ind); }

thanks