CLIJ using 100% on all CPU Cores when applying Weka Model

Hi All,

Sorry @haesleinhuepf another CLIJ/Weka question.

I have it all working like. I have trained a weka model using CLIJ building feature stacks and training is insanely fast. When it goes to apply that model though it is slamming att 12 CPU cores to 100% for a couple of minutes. The result comes out fine but it seem odd to have such a high CPU load on something that should be handed off to the GPU.

These are big images. The feature stacks are about 3GB but i am sending them to a 3090 with 24GB on it so it should be ok i thought and not need to overflow to somewhere else.

In fact, CLIJ only accelerates the feature stack generation. The underlying Weka-classification uses the original CPU-based code.

If your images become bigger, you can also try the GPU-accelerated version of labkit. I’m not sure if it can also be used from macro already. Maybe @maarzt can comment on that.

Let us know if this helps :slight_smile:


Ah OK, that makes sense then. It’s scraping into the 32gb ram of my laptop so it’s ok for now. Final version will be ran on a workstation with 128gb ram and an rtx titan so we should be safe.

1 Like