Hi, have anyone compared the GPU time spent on NMS in caffe and pytorch? I recently transfer a model from caffe to libtorch. And I found that the main time spent by libtorch is concentrated on the NMS operation. the time is 160ms, which takes 80% of all forward time. but the overall time consumption…

Yeah, well, the usual way to do these things is to grab an input and try to measure the function in isolation. Benchmarking has quite a few pitfalls, in particular with CUDA asynchronous computation involved, so it’s hard to say whether you found something where PyTorch is indeed terribly slow or w…

tom (Thomas V) July 26, 2019, 9:22am 9

Any analog of torch.cuda.syncronize() in C++ API? C++

Hi, You can call directly the cudaDeviceSynchronize() method as is done by the python code here.

show post in topic

Home
Categories
Guidelines
Terms of Service
Privacy Policy

Powered by Discourse, best viewed with JavaScript enabled

Nms gpu cost time so more than caffe