model inference in docker container

[noway]

hi,  community

# iMX 8MP

@ Harvey021

I have encountered an issue regarding model inference—it fails inside the Docker container but works on the host machine. The relevant shared libraries have already been copied into the container, and the program runs without reporting any missing dependencies

When I ran the official verification program, it encountered a segmentation fault.

Execute the command as follows：

root@imx8mpevk:~/examples# ./label_image --external_delegate_path=/usr/lib/libvx_delegate.soo
INFO: Loaded model ./mobilenet_v1_1.0_224_quant.tflite
INFO: resolved reporter
INFO: Vx delegate: allowed_cache_mode set to 0.
INFO: Vx delegate: device num set to 0.
INFO: Vx delegate: allowed_builtin_code set to 0.
INFO: Vx delegate: error_during_init set to 0.
INFO: Vx delegate: error_during_prepare set to 0.
INFO: Vx delegate: error_during_invoke set to 0.
INFO: EXTERNAL delegate created.
INFO: Applied EXTERNAL delegate.
W [HandleLayoutInfer:332]Op 162: default layout inference pass.
Segmentation fault

gdb debug:

docker image: ubuntu24.04-arm64

docker command: docker run -it --rm --device=/dev/galcore --device=/dev/mxc_hantro --device=/dev/mxc_hantro_vc8000e --network=host --privileged=true ubuntu:dl_gdb bash

target image version: linux6.6.36 full

target machine: i.MX 8mp

[Chavira]

Hi @noway, 

I hope you are doing very well.

Please take a look to the chapter 2.6.2 Building the TensorFlow Lite Library with the Flex Delegate for i.MX Linux platforms and 2.6.2.2 Setting up Docker VM of i.MX Machine Learning User's Guide.

I hope this can helps.

Best regards,

Chavira

[noway]

Thank you for your answer. But I was running Docker on 8mp. My base image is arm64/v8/ubuntu24.04