This model is a causal language model for the Italian language, based on a GPT-like [1] architecture (more specifically, the model has been obtained by modifying Meta’s XGLM architecture [2] and exploiting its 7.5B pre-trained checkpoint as a starting point). The model has ~6.6B parameters and a vocabulary of 50.335 tokens. It’s an instruction-based model, trained with low-rank adaptation, and it’s mainly suitable for general purpose prompt-based tasks involving natural language inputs and outputs. The current release is a first prototype of this model, and it will be upgraded over time. The released checkpoint is quantized in 8-bit, so that it can easily be loaded and used for training and inference on ordinary hardware like consumer GPUs, and it requires the installation of the transformers library version >= 4.30.1 and the bitsandbytes library, version >= 0.37.2 On Windows operating systems, the bitsandbytes-windows module also needs to be installed on top. However, it appears that the module is not yet updated with some recent features, like the possibility to save the 8-bit quantized models. In order to include this, you can install the fork in this repo, using: The model has…
Модальности:
Генерация текста
Задача: Генерация текста
Автор: osiria
Теги: xglm, it, endpoints_compatible, 8-bit
Лайков: 2 | Загрузок: 121
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
