Segmentation fault when doing context.enqueue

My network is based on tensorflow with plugin layer and i convert it into UFF converter. The stage of uff parser and serialize are all success,but when I call the context.enqueue it happens “Segmentation fault”. And I found the time occurs fault is that when the first plugin layer begin using context get debug function. I wonder what can be the reason of this situation?

Hello, Can you provide the traceback or log? Also, can you provide details on the platforms you are using?

Linux distro and version
GPU type
nvidia driver version
CUDA version
CUDNN version
Python version [if using python]
Tensorflow version
TensorRT version

can you also post any logs/call tracebacks from segmentation fault? Also, posting your source will help us debug it as well.

My gpu type is Tesla P4 and my environment is ubuntu 16.04,tensorRT4,cuda9.0,cudnn7.1,python 2.7. My log is a little long. But it said that when the context execute to my first plugin layer it occurs segmentation fault. It is normal during the parse uff stage.

Thanks, can you also post the segmentation fault traceback? It will help us debug your issue.

Here is some of my log, _ResizeBilinear1 is my plugin layer.
INFO: Debug synchronize completed successfully after execute: tower_0/resnet_v1_50_4/block4/unit_3/bottleneck_v1/conv2/Conv2D + tower_0/resnet_v1_50_4/block4/unit_3/bottleneck_v1/conv2/Relu (type=3)
INFO: About to execute layer: tower_0/resnet_v1_50_4/block4/unit_3/bottleneck_v1/conv3/Conv2D + tower_0/resnet_v1_50_4/block4/unit_3/bottleneck_v1/add + tower_0/resnet_v1_50_4/block4/unit_3/bottleneck_v1/Relu (type=14)
INFO: Debug synchronize completed successfully after execute: tower_0/resnet_v1_50_4/block4/unit_3/bottleneck_v1/conv3/Conv2D + tower_0/resnet_v1_50_4/block4/unit_3/bottleneck_v1/add + tower_0/resnet_v1_50_4/block4/unit_3/bottleneck_v1/Relu (type=14)
INFO: About to execute layer: tower_0/lateral/res5/Conv2D + tower_0/lateral/res5/Relu (type=2)
INFO: Debug synchronize completed successfully after execute: tower_0/lateral/res5/Conv2D + tower_0/lateral/res5/Relu (type=2)
INFO: About to execute layer: _ResizeBilinear1 (type=12)
Segmentation fault (core dumped)

Thanks for the log (which indicates everything is ok until _ResizeBilinear1 layer. Is there anyway for you to get the core dump? (specifically the traceback)

For example:


then, issue bt command.