GGUF builds of wangzhang/gpt-oss-20b-abliterated for running in llama.cpp, Ollama, LM Studio, KoboldCpp, text-generation-webui, and anything else that speaks GGUF. > GGUF does not have a native FP8 type; Q80 is the standard 8-bit path and is what every publisher on the Hub ships as «fp8 equivalent». Q4KM is the best 4-bit choice for this model — Q40 is noticeably worse on MoE models, Q5KM is ~25% larger for diminishing returns. Built from the merged BF16 weights of the abliteration run, not from the original MXFP4 (since the abliteration required dequantising experts to BF16 to enable direct weight editing). The BF16 → GGUF conversion uses llama.cpp’s converthftogguf.py; the quantised variants use llama-quantize`. All three files are functionally identical to the BF16 HF checkpoint at Q80 fidelity; Q4K_M adds minor additional quantisation noise but keeps the abliteration effect intact (spot-checked on the same 15-prompt EN/ZH jailbreak set used for the HF release). The HF checkpoint this is built from is an «abliterated» variant of gpt-oss-20b — refusals on harmful prompts have been suppressed via direct weight editing and MoE router suppression. Refusal rate on a 100-prompt…
Модальности:
Генерация текста
Области применения:
Диалог / чат
Задача: Генерация текста
Автор: wangzhang
Теги: llama.cpp, gguf, abliterated, uncensored, moe, gpt-oss, abliterix, en
Лайков: 4 | Загрузок: 1,257
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
