search

CMU-Perceptual-Computing-Lab / caffe_rtpose

Realtime C++ code for multi-person pose estimation
Other
356 stars 207 forks source link

F0406 17:19:19.386066 12837 syncedmem.cpp:56] Check failed: error == cudaSuccess (2 vs. 0) out of memory #36

Closed cheneyfan closed 7 years ago

NHZlX commented 7 years ago

I also meet the same problem(i use the nvidia tesla graphics with 12G memory). Have u solved it?

gineshidalgo99 commented 7 years ago

You must use cuDNN to avoid out of memory. And If you are using it, maybe you could also reduce the batch size.

NHZlX commented 7 years ago

@gineshidalgo99 thanks, it works.