Generally looks good to me.

There is one change that I think should be made (assuming I've understood it correctly). In generate_config_file.py, I think the bounds file represents the GEE image downloads, not the chips for inference - Is that right? If so, then the variable name chips_gdf is misleading. I've changed this to bounds_gdf. I've also tweaked the CLI parameter description to remove the word "chips".

I've added a small commit with these changes. Feel free to revert my change if I've got it wrong.

Thank you @andrewphilipsmith, this makes sense! I'll fix the broken test and merge it.