segnet-console and Jetson Nano examples

Autonomous Machines Jetson & Embedded Systems Jetson Nano
1

I can run imagenet-console and imagenet-camera fine (commit f19a2a78b6d3a563dd81750d65582651a161684e) . But if I run the

./segnet-console peds-001.jpg segs.jpg

in the Jetson Nano, the program will hang in builder->buildCudaEngine() (waited > 15 minutes).

I get the usual

[TRT]  TensorRT version 5.0.6
[TRT]  detected model format - caffe  (extension '.caffemodel')
[TRT]  desired precision specified for GPU: FASTEST
[TRT]  requested fasted precision for device GPU without providing valid calibrator, disabling INT8
[TRT]  native precisions detected for GPU:  FP32, FP16
[TRT]  selecting fastest native precision for GPU:  FP16
[TRT]  attempting to open engine cache file networks/FCN-Alexnet-Cityscapes-HD/snapshot_iter_367568.caffemodel.2.1.GPU.FP16.engine
[TRT]  cache file not found, profiling network model on device GPU
[TRT]  device GPU, loading networks/FCN-Alexnet-Cityscapes-HD/deploy.prototxt networks/FCN-Alexnet-Cityscapes-HD/snapshot_iter_367568.caffemodel
[TRT]  retrieved Output tensor "score_fr_21classes":  21x26x58
[TRT]  retrieved Input tensor "data":  3x1024x2048
[TRT]  device GPU, configuring CUDA engine
[TRT]  device GPU, building FP16:  ON
[TRT]  device GPU, building INT8:  OFF
[TRT]  device GPU, building CUDA engine (this may take a few minutes the first time a network is loaded)

The code hangs on tensorNet.cpp:537 with the

nvinfer1::ICudaEngine* engine = builder->buildCudaEngine(*network);

I can also see that there is a increasing amount of ioctl-calls with strace

ioctl(5, HIDIOCSFEATURE(16), 0x7fc596c210) = -1 EAGAIN (Resource temporarily unavailable)
ioctl(5, HIDIOCSFEATURE(16), 0x7fc596c210) = -1 EAGAIN (Resource temporarily unavailable)
ioctl(5, HIDIOCSFEATURE(16), 0x7fc596c210) = -1 EAGAIN (Resource temporarily unavailable)
ioctl(5, HIDIOCSFEATURE(16), 0x7fc596c210) = -1 EAGAIN (Resource temporarily unavailable)
ioctl(5, HIDIOCSFEATURE(16), 0x7fc596c210) = -1 EAGAIN (Resource temporarily unavailable)
ioctl(5, HIDIOCSFEATURE(16), 0x7fc596c210) = -1 EAGAIN (Resource temporarily unavailable)
ioctl(5, HIDIOCSFEATURE(16), 0x7fc596c210) = -1 EAGAIN (Resource temporarily unavailable)
ioctl(5, HIDIOCSFEATURE(16), 0x7fc596c210) = -1 EAGAIN (Resource temporarily unavailable)
ioctl(5, HIDIOCSFEATURE(16), 0x7fc596c210) = -1 EAGAIN (Resource temporarily unavailable)
ioctl(5, HIDIOCSFEATURE(16), 0x7fc596c210) = -1 EAGAIN (Resource temporarily unavailable)
ioctl(5, HIDIOCSFEATURE(16), 0x7fc596c210) = -1 EAGAIN (Resource temporarily unavailable)
ioctl(5, HIDIOCSFEATURE(16), 0x7fc596c210) = -1 EAGAIN (Resource temporarily unavailable)
ioctl(5, HIDIOCSFEATURE(16), 0x7fc596c210) = -1 EAGAIN (Resource temporarily unavailable)
ioctl(5, HIDIOCSFEATURE(16), 0x7fc596c210) = -1 EAGAIN (Resource temporarily unavailable)
ioctl(5, HIDIOCSFEATURE(16), 0x7fc596c210) = -1 EAGAIN (Resource temporarily unavailable)

Any thoughts?

Thanks!

2

Hi,

Thanks for reporting this.
We will try to reproduce this issue on our environment first.

Thanks.

3

Hi,

We can execute this sample without any error but it really takes a long time. (around 34min)
Please help to test if you can also get the sample work after a longer waiting time.

The most execution time is used for compiling a TensorRT engine.
Due to the limited resource, the compiled time is pretty long on Jetson Nano.

But please noticed that you only need to apply the compilation one time.
TensorRT can be launched with the serialized engine afterward.

Thanks.

4

Patience indeed was the key!
Thank you!