DeepSeek-R1-Distill-Qwen-25.5B with Brainstorm 40x, (88 layers, 1043 tensors) Required: CHATML template. ( OR standard «Jinja Autoloaded Template» (this is contained in the quant and will autoload) ) Keep in mind this model is experimental and may require one or more regens to work, especially with the «think» system of Deekseek involved here. Brainstorm 40x is by DavidAU, and extends the «decision making» and «creativity» of an LLM/AI. Higher temps will result in deeper, richer «thoughts»… and frankly more interesting ones too. The «thinking/reasoning» tech (for the model at this repo) is from the original Qwen 2.5 «Distill» model from Deepseek: [ https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-14B ] In this case, Brainstorm 40x module was grafted directly onto «DeepSeek-R1-Distill-Qwen-14B» bringing it up to 88 layers, 25.5B parameters. This version will retain all the functions and features of the original «DeepSeek» model. Please visit their repo for all information on features, test results and so on. This model is for all use cases, and it has a slightly more creative slant than a standard model. This model can also be used for solving logic puzzles, riddles,…
Модальности:
Генерация текста
Области применения:
Логика и рассуждение Диалог / чат
Задача: Генерация текста
Автор: DavidAU
Теги: gguf, All use cases, reasoning, thoughts, deep thinking, deepseek, creative, creative writing
Лайков: 4 | Загрузок: 422
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
