Hey, I never claimed it was going to be performant, did I? That would be a lie. I am only focused on making my code readable and usable. NOT assuming people to be ML experts and expect people to know what they are doing. There’s a reason to why your “free” option didn’t get mass adoption compared to DataPredict and Kironte’s Neural Network. I think you should have studied why LuNet, RoTorch, EasyML, OpenML and other ML frameworks fail to gain traction. (Hint: It is the same reason as yours: they don’t want to learn a framework, they want to use it.)
If you’re going to mock me (that lol is a dead give away) like that, I’m just going to keep my mouth shut about certain “implementations” of yours and let you ruin your own credibility. Meanwhile, I have three years worth of mistakes and debugging them to see what issues I have created and fix my misunderstanding.
You should expanded more on “what template is”, because you could be referencing anything! Sword fighting AI template? A sample reference code for using DQL?
Also, I do not care about performance if your “correctness” have a problem. And do you know what’s the issue with OpenML’s implementation? Gradient (not Gradien) isn’t even being implemented in CNN and his DQL has implementation issues. You can keep saying OpenML is much simpler, but that means nothing if the math implementation is incorrect.
Claiming OpenML to be “performant” while ignoring his implementation problems makes me question your knowledge of ML, DL and RL in general. After all, I have been doing this for three years.
Plus, I’m keeping it in pure Lua since I want it to grow outside of Roblox ecosystem, not stuck inside it.