search

CaoWGG / TensorRT-CenterNet

tensorrt5 , centernet , centerface, deform conv, int8
767 stars 158 forks source link

推理出现：段错误 (核心已转储) #66

Open tms2003 opened 4 years ago

tms2003 commented 4 years ago

在使用ｃｅｎｔｅｒｆａｃｅ出错，命令如下： ./runDet -e /home/tms/source/TensorRT-CenterNet/model/centerface.engine -i /home/tms/source/opencvdemo/test.jpg -c /home/tms/source/TensorRT-CenterNet/test.h264

    void ctdetNet::doInference(const void *inputData, void *outputData)
    {
        const int batchSize = 1;
        int inputIndex = 0 ;
        CUDA_CHECK(cudaMemcpyAsync(mCudaBuffers[inputIndex], inputData, mBindBufferSizes[inputIndex], cudaMemcpyHostToDevice, mCudaStream));
        mContext->execute(batchSize, &mCudaBuffers[inputIndex]);

我的环境是：TensorRT-6.0.1.8，CUDA10.2,我跟踪了你的代码出错是在CUDACHECK这句，请问可能是什么原因呢？

KevenLee commented 4 years ago

在使用ｃｅｎｔｅｒｆａｃｅ出错，命令如下： ./runDet -e /home/tms/source/TensorRT-CenterNet/model/centerface.engine -i /home/tms/source/opencvdemo/test.jpg -c /home/tms/source/TensorRT-CenterNet/test.h264

    void ctdetNet::doInference(const void *inputData, void *outputData)
    {
        const int batchSize = 1;
        int inputIndex = 0 ;
        CUDA_CHECK(cudaMemcpyAsync(mCudaBuffers[inputIndex], inputData, mBindBufferSizes[inputIndex], cudaMemcpyHostToDevice, mCudaStream));
        mContext->execute(batchSize, &mCudaBuffers[inputIndex]);

我的环境是：TensorRT-6.0.1.8，CUDA10.2,我跟踪了你的代码出错是在CUDACHECK这句，请问可能是什么原因呢？

This means something wrong with memory! you should modify include/ctdetConfig.h to fix it

Jumponthemoon commented 4 years ago

在使用ｃｅｎｔｅｒｆａｃｅ出错，命令如下： ./runDet -e /home/tms/source/TensorRT-CenterNet/model/centerface.engine -i /home/tms/source/opencvdemo/test.jpg -c /home/tms/source/TensorRT-CenterNet/test.h264

    void ctdetNet::doInference(const void *inputData, void *outputData)
    {
        const int batchSize = 1;
        int inputIndex = 0 ;
        CUDA_CHECK(cudaMemcpyAsync(mCudaBuffers[inputIndex], inputData, mBindBufferSizes[inputIndex], cudaMemcpyHostToDevice, mCudaStream));
        mContext->execute(batchSize, &mCudaBuffers[inputIndex]);

我的环境是：TensorRT-6.0.1.8，CUDA10.2,我跟踪了你的代码出错是在CUDACHECK这句，请问可能是什么原因呢？

请问解决了吗？

sailyung commented 4 years ago

you should modify include/ctdetConfig.h to fix it, and then re build cmake

Jumponthemoon commented 4 years ago

you should modify include/ctdetConfig.h to fix it, and then re build cmake

Works perfect!Thanks!

tms2003 commented 4 years ago

ok,this work for my trt 5 ,but not result show in trt 6

yezechen commented 4 years ago

you should modify include/ctdetConfig.h to fix it

how to fix it？