This is a special release of Dolphin-DPO based on the LASER paper and implementation by @fernandofernandes assisted by @ehartford We have further carried out a noise reduction technique based on SVD decomposition. We have adapted this paper on our own version of LASER, using Random Matrix Theory (Marchenko-Pastur theorem) to calculate optimal ranks instead of brute-force search. This model has achieved higher scores than 2.6 and 2.6-DPO. Theoretically, it should have more robust outputs. This model is uncensored. I have filtered the dataset to remove alignment and bias. This makes the model more compliant. You are advised to implement your own alignment layer before exposing the model as a service. It will be highly compliant to any requests, even unethical ones. Please read my blog post about uncensored models. https://erichartford.com/uncensored-models You are responsible for any content you create using this model. Enjoy responsibly. It took 3 hours to tune the model on SVD rank reduction on a RTX 4090 24 GB of RAM, following our Marchenko-Pastur approach. Prompt format: This model uses ChatML prompt format. NEW — maps to tokenid 2. This is the same tokenid as so…
Модальности:
Генерация текста
Задача: Генерация текста
Автор: LoneStriker
Теги: mistral, en, text-generation-inference, endpoints_compatible
Лайков: 3 | Загрузок: 16
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.