Super-Slop-Machina-Macro-Roleplay-Instruct-350M-0812: Local Roleplay, Now Coming to Your Fridge.
5467740410491902164
"Smaller than ever before, now runnable anywhere, no matter whether you prefer running your model on toaster or on the fridge!."
Improvements:
Better at roleplay than the base model
Writes more consistently
Somewhat better at keeping track of the plot
Handles long term conversations slightly better now
Quants:
Are lying in the repo, I can't say anything for sure about them but wouldn't recommend going any lower than Q6_K
Training and dataset details:
I specifically used my own private dataset consisting of my best chatlogs for this finetune, method used was full parameter finetune with high LR, resulting in drastic change to how the model writes and expresses itself, as I mentioned in many of my previous models of such scale, do not expect it to suddenly feel like its R1 you're talking with, this model is TINY.
Big announcement:
I experimented with Qwen 3.5 9b and I think I am somewhat succeeding, I will soon release the finetuned weights, result that I was able to achieve with it was VERY interesting...
