is there a way to control optimization and compression levels?
What are the defaults used by hailomz?
Is there a way to disable compression?
You can control both from the allocation script, all the details are in the user guide. Generally speaking those are the commands:
model_optimization_flavor(optimization_level=4)
model_optimization_flavor(compression_level=2)
The default depends behaviour depends on the model size, and system configuration (If you have a supported GPU or not).