oN8WeJbq5Jo7Pk/CVE-Heretic-Fable5-27B-GGUF - Каталог нейросетей
Генерация текста

oN8WeJbq5Jo7Pk/CVE-Heretic-Fable5-27B-GGUF

Добавлено:
oN8WeJbq5Jo7Pk/CVE-Heretic-Fable5-27B-GGUF

A Qwen3.6-27B finetune for CVE writeups and reverse engineering. Abliterated. Trained on a curated, source-balanced 28.6K-row SFT mix in which CVE writeups, distilled reasoning, and agentic traces each contribute meaningful weight (cap of 12K rows per source so no single dataset dominates). Then abliterated to remove refusals on lawful security research. The LoRA adapter is merged back into the base before quantization, so this ships as a single standalone GGUF. — Base: Qwen/Qwen3.6-27B (Apache 2.0) — Recipe: LoRA r=32, alpha=64, targetmodules=all-linear on 8× B200 (DDP), constant LR + warmup, eval-driven early-stopping with patience 4, length-grouped sampler. Adapter then merged into base. — Abliteration: 1000-trial Optuna sweep using the Heretic recipe; trial 93 selected from the Pareto front of fewest refusals and lowest KL. KL = 0.0041, refusals = 10/100 down from a baseline of 90/100, with the rest of model behavior essentially unchanged. — Quantization shipped: Q5K_M GGUF (~18 GB). Fits cleanly on a single 24 GB GPU (3090/4090) at 32K context, 16 GB with KV-cache quantization. This is a Qwen3.6 thinking-mode-capable model. Recommended sampling (matches what I trained and…

Модальности:
Генерация текста

Области применения:
Генерация кода Диалог / чат


Задача: Генерация текста
Автор: oN8WeJbq5Jo7Pk
Теги: gguf, qwen3.6, merged, abliterated, heretic, security, cve, reverse-engineering
Лайков: 4  |  Загрузок: 296

Открыть на HuggingFace →

Описание основано на материалах HuggingFace. Перевод выполнен автоматически.