python-semver VS system-intelligence

To do this I will use the package python-semver to easily handle it. I'll define a function that I can run on the current save data version, and the new code data version, to say whether this upgrade is safe or not:

mlf-core provides CPU and GPU deterministic project templates. Hence, when you are using a mlf-core template you can be sure that you always get the bit exact same results given the same hardware. mlf-core ensures this by tracking all parameters and metrics with MLflow, tracking all hardware with system-intelligence, containerizing the environment with Conda and MLflow and the final spicy ingredient: mlf-core lint. This custom static code analyzer evaluates your code for two things: 1. You are forcing deterministic algorithms. Pytorch and Tensorflow use non-deterministic algorithms by default, but there ways to force part of them to behave 2. You are NOT using non-deterministic algorithms. If any of those are found mlf-core lint will alert you and tell you which function in which file and line violates determinism. You can then replace this method with a deterministic workaround.