GPU not utilized by Hailo python Optimize() API

General
,
1

I’m using Hailo DFC in a custom docker container.

I tried using hailo’s command line tool to run optimization (hailo optimize) on a parsed model. I observed that the GPU is utilized in this stage.

However, when I try to optimize the same parsed model (with same model file and calibration dataset) using hailo’s python API (runner.optimize()), the optimization is happening on CPU and not GPU.

Since the command line tool is utilizing GPU, it is safe to assume that the container has access to the GPU and all necessary drivers are installed.

Are there any flags/configurations related to hailo’s python APIs that can cause this behaviour?

Regards,

Shashank

2

We do provide the Hailo AI Software Suite Docker in the Developer Zone.

I do not think so. Did you get any warning messages? e.g.

[warning] Reducing optimization level to 0 (...) because there's no available GPU