— Type: decoder-only — Tokenizer: BPE — Vocabulary size: 32000 — Max sequence length: 120 — Hidden size: 768 — FFN size: 3072 — Attention heads: 24 — Decoder layers: 4
Модальности:
Генерация текста
Задача: Генерация текста
Автор: ilnikolaev
Теги: tflite, ru
Лайков: 2 | Загрузок: 9
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
— Simple and Clear: MINIMAL code and architecture changes. Only one up-and-down projection layer introduced, NO recurrent memory...
In addition to pytorch and transformers, install required packages: mediocredev/open-llama-3b-v2-chat is based on LLaMA 3B v2. It can...
This is a merge of pre-trained language models created using mergekit. This model was merged using the SCE...
An OpenLlama-3B finetune for role-play, instruction following, and code generation For a guide on how this was fine...
