Check failed: error == cudaSuccess (2 vs. 0) out of memory

[jxylon]

I run demo to extract bounding box features on GTX 2080Ti, but I received this error.

F0115 10:36:21.001302  3456 syncedmem.cpp:71] Check failed: error == cudaSuccess (2 vs. 0)  out of memory

Looking forward to anyone's help

[zBabar]

Hi,

Have you solved this problem? I am facing similar problem for some images. Rest of the images it do work. But some images it gets stuck and throws this error. F0324 13:21:55.416903 11805 syncedmem.cpp:71] Check failed: error == cudaSuccess (2 vs. 0) out of memory

[maoyj1998]

I run demo to extract bounding box features on GTX 2080Ti, but I received this error.

F0115 10:36:21.001302  3456 syncedmem.cpp:71] Check failed: error == cudaSuccess (2 vs. 0)  out of memory

Looking forward to anyone's help

hi have you solved this problem ?

[maoyj1998]

Decreasing BATCH_SIZE and RPN_BATCHSIZE size in yml configuration file fixed it for me (experiments/cfgs/faster_rcnn.. .yml)

hi, I check the yml file, but all the config you mentioned are training settings, test settings have nothing todo with BATCH_SIZE, I wonder whether it will work if I change these config

[maoyj1998]

Hi,

Have you solved this problem? I am facing similar problem for some images. Rest of the images it do work. But some images it gets stuck and throws this error. F0324 13:21:55.416903 11805 syncedmem.cpp:71] Check failed: error == cudaSuccess (2 vs. 0) out of memory

hi, did you solve this problem?

[SandroJijavadze]

@maoyj1998 I deleted my previous comment, I thought it was fixed but I still got the same error after running on few examples. I switched to T4 GPU wchich has 16gb of memory and that fixed it later. I was getting errors on rtx 2070 super which has only 8gb memory.

[maoyj1998]

@maoyj1998 I deleted my previous comment, I thought it was fixed but I still got the same error after running on few examples. I switched to T4 GPU wchich has 16gb of memory and that fixed it later. I was getting errors on rtx 2070 super which has only 8gb memory.

I use this code to extract feature from VG datasets, and I found it was caused by some images with big difference in aspect ratio, for example 281 * 500, faster rcnn will resize the image based on the shorter edge, so making the larger edge too large.
I modified this part and it works.

[frozen30]

Have you solved this problem? I will also encounter this problem when running demo.ipynb