70Blivion v0.1 is a model in the release candidate stage, based on a merge of L3.1 Nemotron 70B & Euryale 2.2 with a healing training step. Further training will be needed to get this model to release quality. This model is designed to be suitable for creative writing and roleplay. This RC is not a finished product, but your feedback will drive the creation of better models. This is a release candidate model. It has some known issues and probably some unknown ones too, because the purpose of these early releases is to seek feedback. — The model is still somewhat censored, though much less than Nemotron 70B. A jailbreak prompt can be used to decensor it further. — The model likes to reproduce parts of the system prompt in its output. This can be reduced by prompting against it, but there will be swipes/edits. — The model can shy away from NSFL content, though I have found that a system prompt can mitigate this avoidance pretty easily. FP16 Safetensors Static GGUF — thanks to mradermacher * IMatrix GGUF — thanks to mradermacher My Discord server — requires Discord (preferred). The Community tab — requires HF login. The SillyTavern Discord thread — must be on SillyTavern Discord. …
Модальности:
Генерация текста
Области применения:
Диалог / чат
Задача: Генерация текста
Автор: invisietch
Теги: llama, not-for-all-audiences, axolotl, qlora, conversational, en, text-generation-inference, endpoints_compatible
Лайков: 4 | Загрузок: 17
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
