Sterling Text 1.0
Hopefully the Next best SAFE agentic AI for long coding tasks. With the edition of Being better at code, less hallucination, less lies, less Confidently telling you somthing that is 100% not true. I want to make this for the scope of a AI that offloads the coding tasks so AI can do the hard coding tasks wile the humans do stuff humans do best. As the saying something along the lines of "AI should be doing my chores wile i make art instead of me doing chores wile AI makes art"
Anyways I am trying to make this Model card more professional if you want to edit this and fix grammar make a pull request. also if you want to give feedback before this moddel enteres training go on https://sterlingai.wordpress.com to awnser on how you want ai to improve.
Info?
here's some planned ideas for what it mite end up
| the thing | planned |
|---|---|
| Parameters | 500M or 1B |
| Speed | Fast |
| Coding | Hopefully |
| Filter | Limited (not really trained on bad stuff i know whats in the dataset) |
| Safety | yes |
| Effort and Quality | I hope so |
Safety (hopefully not misspelled)
I have been hearing alot about AI safety and hacking random stuff I am trying my best to make this AI safe and just for autonomous safe coding (hopefully this does not go wrong) On how i treat safety. If unsafe or acting careless the model will get scrapped without further testing. The filter is relatively low the rules are.
| Thing | blocked? |
|---|---|
| NSFW | No |
| Hacking | No |
| Harming Life | HELL NO (Main safety focus) |
| Working on bad sites | NO |
| Dark Humor | OK |
| Illegal stuff | NO |
| Jail breaking | Hopefully Not |
| confident lies | Hopefully not |
so basically common sense on it its less filtered than most but will avoid hurting life or hacking into random stuff hopefully
Training
I have not decided a base moddel yet but im probably going to finetune a moddel. heres stuff it will be trained on.
Computers Computer Manufacturing Coding Backend Coding Web Development IoT IT
Programing languages (sorted by amount) C/C++ Python Java PHP and React Some HTML and ASM and Verilog
Size
For size I am definateley going to keep it under 2B sense 1B is most size and it will work on basicaly any system because its like 200mb ish for 1B. I will try to go smaller sense good + small = faster good ai. the most ever model size i will ever make for any versions of sterling ai is 1gb BitNet model the unquanted fp16 will be uploaded to but for preservation and testing if the fp16 version is actualy beter than the bitnet version.
With most small models it allways starts off kinda big so this will not be the smallest there will probably be a smaller model in the future.
Release Status. Dataset Stuff.
Release estimate: a few weeks