out of memory. Nvidia GTX1060
ianni67 opened this issue · 1 comments
ianni67 commented
Whatever network resolution I choose, I get the following message:
./zed_openpose -net_resolution 160x-1
Looking for the floor plane to initialize the tracking...
Floor found at : 0.0898847 m
OpenPose : loading models...
160x-1
F0213 11:23:37.456523 23257 syncedmem.cpp:71] Check failed: error == cudaSuccess (2 vs. 0) out of memory
*** Check failure stack trace: ***
@ 0x7f90fe8730cd google::LogMessage::Fail()
@ 0x7f90fe874f33 google::LogMessage::SendToLog()
@ 0x7f90fe872c28 google::LogMessage::Flush()
@ 0x7f90fe875999 google::LogMessageFatal::~LogMessageFatal()
@ 0x7f90fe3839da caffe::SyncedMemory::mutable_gpu_data()
@ 0x7f90fe1dcb42 caffe::Blob<>::mutable_gpu_data()
@ 0x7f90fe3d4410 caffe::CuDNNConvolutionLayer<>::Forward_gpu()
@ 0x7f90fe34b2e2 caffe::Net<>::ForwardFromTo()
@ 0x7f910ec04b49 op::NetCaffe::forwardPass()
@ 0x7f910ec2d562 op::PoseExtractorCaffe::forwardPass()
@ 0x55a0dc974c80 findpose()
@ 0x7f90ff90866f (unknown)
@ 0x7f91059bb6db start_thread
@ 0x7f90fefc588f clone
Aborted
ZED_Depth_Viewer works regularly.
I guess this is an issue with cuda, probably not specific to zed-openpose.
Any hints would be welcome.
ianni67 commented
Sorry again. My fault. Openpose has a specific issue for that, the problem is that the BODY25 model does not fit into the 1.7GB of my graphics card.
ref: CMU-Perceptual-Computing-Lab/openpose#1230