Varying hardware configurations

[matteoettam09]

Simulate many different types of hardware configurations.

Less generically, introduce different kind of "common" hardwares. This means selecting the most common set of combinations of the components of the state space associated with hardware (this means we need all possible hardwares to have these variables): https://pcpartpicker.com/

Once defined the variables characterizing the state (CPU, GPU, RAM,..). From the legacy:

machine_attributes = {"CPU", "RAM", "GPU", "created_at", "timeout"}

Then, you can build a set of dictionaries for digital twins:
{raspberrypi400: {RAM: 1, CPU: 4, ?:?, ?:?, ...}, xx:{...}} to upload into the agent class.
In real word, you have IOT, Mobile Phones, etc.

Hardware profiling/understanding of internal state: Use programs to load proper hardware profile ones moving away from synthetic profiles.

See e.g.: https://github.com/zama-ai/concrete-ml/blob/bc01ad187c38997a904156f3a2496982e2941a54/benchmarks/convert_cifar.py#L50