---
license: mit
---
*this version cuda optimize colab friendly json version of handcraft parameters model*
weights input: "weights.json" file 

100% accuracy on 10 million random test cases w/ only 343 parameters. 
As a bonus, it uses the vanilla Qwen3 architecture, just with the right weights.

*credit X.com*
https://x.com/N8Programs/status/2026318679525163189

*credit github.com*
https://gist.github.com/N8python/02e41d156ec615328cde2e1e5c0e9d53

*credit for me*

my name is : Ashen pashidu

I'm working with colab projects 
I'm a student 

*warning*
This is not my idea. I'm do that's code optimize for colab & weights input method change 
this original idea is https://x.com/N8Programs/status/2026318679525163189