Make a custom qwen 3.8 27B abliterated for ninfer?

A user running ninfer's Qwen 3.8 27B model on an RTX 3090 is seeking guidance on creating an abliterated version of the model, noting that ninfer uses a custom file format that prevents simple configuration changes. The user is exploring whether porting an abliterated Qwen model to ninfer's custom inference engine is feasible or if building a custom inference solution would be easier.

I just got ninfer qwen 3.8 27B working on my 3090. It works really well but I want to experiment with more freedom on an abliterated model. Can anyone tell me how this could be achieved? It looks like it could be easy, or very hard. So far I know it’s not a config change. The ninfer is a custom file format that I don’t know how it works. I haven’t read much yet but I thought I’d ask as I continue testing regular ninfer qwen out and studying how an abliterated model might get ported to a custom inference thing. I don’t even know if making a custom inference thing would be easier than porting the ablit qwen.