Comment on: Support native NVFP4 / ModelOpt checkpoints (e.g. Qwen3.5-9B-NVFP4)
Repo: elder-plinius/OBLITERATUS by Vastopian
I don't think you abliterate on quant models. I'm pretty sure you need to abliterate first then quant to nvfp4. I think it only uses 4bit for the finding the refusals.
GitHub Issue
Parent Entity
State: Open • Comments: 2
Other Comments / Reviews
SaaS Metrics