Failed to run detectnet-console with DLA_0 device

Jetson, OS:
Linux jetson-0423318033018 4.9.108-tegra #1 SMP PREEMPT Fri Sep 28 22:03:31 PDT 2018 aarch64 aarch64 aarch64 GNU/Linu

Describe the problem
When i run the detectnet-console with DLA_0 device, the TRT reported that "[TRT] device DLA_0, failed to build CUDA engine"

Full log:
detectNet – loading detection network model from:
– prototxt ./networks/facenet-120/deploy.prototxt
– model ./networks/facenet-120/snapshot_iter_24000.caffemodel
– input_blob ‘data’
– output_cvg ‘coverage’
– output_bbox ‘bboxes’
– mean_pixel 0.000000
– threshold 0.500000
– batch_size 2

[TRT] TensorRT version 5.0.0
[TRT] desired precision specified for DLA_0: FASTEST
[TRT] requested fasted precision for device DLA_0 without providing valid calibrator, disabling INT8
[TRT] native precisions detected for DLA_0: FP32, FP16, INT8
[TRT] selecting fastest native precision for DLA_0: FP16
[TRT] attempting to open engine cache file ./networks/facenet-120/snapshot_iter_24000.caffemodel.2.0.DLA_0.FP16.engine
[TRT] cache file not found, profiling network model on device DLA_0
[TRT] device DLA_0, loading ./networks/facenet-120/deploy.prototxt ./networks/facenet-120/snapshot_iter_24000.caffemodel
[TRT] retrieved Input tensor “data”: 3x450x450
[TRT] retrieved Output tensor “coverage”: 1x28x28
[TRT] retrieved Output tensor “bboxes”: 4x28x28
[TRT] device DLA_0, configuring CUDA engine
[TRT] device DLA_0, building FP16: ON
[TRT] device DLA_0, building INT8: OFF
[TRT] device DLA_0, building CUDA engine
[TRT] …/builder/cudnnBuilder2.cpp (689) - Misc Error in buildSingleLayer: 1 (Unable to process layer.)
[TRT] …/builder/cudnnBuilder2.cpp (689) - Misc Error in buildSingleLayer: 1 (Unable to process layer.)
[TRT] device DLA_0, failed to build CUDA engine
device DLA_0, failed to load ./networks/facenet-120/snapshot_iter_24000.caffemodel
detectNet – failed to initialize.
detectnet-console: failed to initialize detectNet

How do i to fix it?

Hi,

We can run detectnet-console without error with TensorRT version 5.0.3, which is from JetPack4.1.1.:
https://developer.nvidia.com/embedded/jetpack

Could you also give it a try?

Thanks.

Note that currently the networks AlexNet, GoogleNet, ResNet-50, and LeNet have been validated on DLA for JetPack 4.1.1 DP.
Please refer to the JetPack Release Notes. Do you have GPU fallback enabled?