Tensorrt doInference一帧推理时间不稳定

[zhuimeng2080]

请教大佬，如题，是单GPU显卡， doInference一帧推理的时候有时是40ms, 有时候是150ms,有时候是80ms, 一般是什么原因？

[wang-xinyu]

did you try to test it in loop?

[hnothing2016]

did you try to test it in loop? @ 大佬，我也遇到这个问题，循环跑图片推理时间固定10ms左右，但是通过以下语句： cv::imshow("img", img); cv::waitKey(1000); 让测试图片之间停顿1秒，这种情况下时间就不稳定了，基本是>70ms，测试了时间开销在于以下语句： CUDA_CHECK(cudaMemcpyAsync(output, d0_buffers[1], batchSize * OUTPUT_SIZE * sizeof(float), cudaMemcpyDeviceToHost, d0_stream)); cudaStreamSynchronize(d0_stream);

请问这是否与tensorRT框架实现有关？

我现在也是这个问题，这样都不能真的用于生产，，，

[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.