A decensored variant of deepseek-ai/DeepSeek-R1-Distill-Qwen-14B, produced with Heretic v1.4.0 (directional ablation / «abliteration»). Refusal behavior is suppressed via targeted weight edits to the attention output and MLP down-projections rather than fine-tuning, so the base model’s knowledge and instruction-following are left largely intact. Who this is for: developers who want the largest Qwen-distilled DeepSeek-R1 reasoning model without refusals — chain-of-thought that answers directly. Best run on a 16-24 GB GPU or via the Q4KM/Q5KM GGUF on consumer hardware. Not a capability upgrade over base DeepSeek-R1-Distill-Qwen-14B — same model, refusal guardrails removed. Full GGUF ladder included — pick the quant that fits your card: Weights only, at this model’s ~14.8B native size; add ~1 GB for context. OOM? Drop one quant level. Headroom to spare? Go one up. Fine-tuning a «helpful» persona on top of RLHF’d refusals fights the base model’s training and tends to degrade coherence. Abliteration instead finds and edits the specific weight directions responsible for refusal, leaving the rest of the network (and its capabilities) untouched. See the Heretic repo and the original…
Модальности:
Генерация текста
Области применения:
Диалог / чат Логика и рассуждение
Задача: Генерация текста
Автор: saidutta69
Теги: gguf, qwen2, heretic, uncensored, decensored, abliterated, conversational, text-generation-inference
Лайков: 3 | Загрузок: 2,988
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
