Spark-X2.5-4B is a compact general-purpose model for conversation, writing, translation, reasoning, coding and tool use with a hybrid attention architecture and up to 1M tokens of native context (English and Chinese). This repository holds three GGUF files of different size made from the BF16 weights with our measured mixed-codec method. Pick one on the Files page; the table below says what each file is. Independent community quantization, not an official XHToken release. Part of the Sakura Mini line. Same method, same texts, same machine for every file; lower KL divergence means closer to the BF16 original: — Sakura-Spark-X2.5-4B-1.71GiB.gguf vs llama.cpp default IQ3XXS (1.68 GiB): KL divergence to the BF16 original 16 % lower on average over all 3 texts (0.03 GiB larger). — Sakura-Spark-X2.5-4B-2.26GiB.gguf vs llama.cpp default IQ4XS (2.21 GiB): KL divergence to the BF16 original 3 % lower on average over all 3 texts (0.04 GiB larger). — Sakura-Spark-X2.5-4B-2.47GiB.gguf vs author’s Q4KM (2.42 GiB): KL divergence to the BF16 original 38 % lower on average over all 3 texts (0.05 GiB larger). Why: the bit budget per tensor is allocated from measured sensitivity (see How it was…
Модальности:
Генерация текста
Области применения:
Диалог / чат
Задача: Генерация текста
Автор: webmp3
Теги: gguf, sakura, sakura-mini, llama.cpp, quantization, imatrix, mixed-quant, sparkx2_5
Лайков: 3 | Загрузок: 376
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
