The code I’m using is similar to the last cell in this notebook, except for the fig.show() part, and it’s looped:
So I run this sequence:
fig = px.choropleth_mapbox() fig.update_layout() fig.write_image(filename)
…hundreds of times in a loop, distributed over several worker processes (one for each CPU). All CPU cores are used, which is good. Nothing is displayed, everything goes direct to disk, I make a bunch of files, then I generate a video with ffmpeg.
My laptop is an Acer Predator Helios 300, which has two GPUs, Intel and NVidia. While the code is rendering the images, I see the Intel GPU being used; there’s zero load on the Nvidia GPU. I suspect the NVidia chip would go faster.
Looks like the orca processes use the GPU.
Is there a way in the code to force using a certain GPU if the system has more than one?