jakeatx/gemma4-12b-mtp-assistant

GGUF conversion of Google’s google/gemma-4-12B-it-assistant draft / MTP assistant model for pairing with google/gemma-4-12B-it-compatible Gemma 4 12B instruction checkpoints. These files are generated for the current Gemma 4 MTP llama.cpp work, not stock mainline builds that reject the assistant architecture. Local runtime validation used am17an/llama.cpp branch gemma4-mtp at commit b8e703e, with a converter compatibility patch so the official Gemma4UnifiedAssistantForCausalLM config registers as the Gemma 4 assistant converter. Requested repo was atx/gemma4-12b-mtp-assistant; the active token is for sjakek and has no atx org write rights, so this package was uploaded under sjakek/gemma4-12b-mtp-assistant. — gemma-4-12B-it-assistant-BF16.gguf: BF16 GGUF, 861,520,128 bytes. — gemma-4-12B-it-assistant-Q80.gguf: Q80 GGUF from branch llama-quantize, 465,109,248 bytes. — logs/: conversion, quantization, load, generation, and long-context validation logs. — Source model: google/gemma-4-12B-it-assistant — Source revision: 723932f88886ab714522468b94f9c7ee48d8c9a8 — Source config architecture: Gemma4UnifiedAssistantForCausalLM — Source model type: gemma4unifiedassistant — GGUF architecture…

Модальности:
Генерация текста

Задача: Генерация текста
Автор: jakeatx
Теги: gguf, gemma4, mtp, speculative-decoding, draft-model, llama.cpp, endpoints_compatible
Лайков: 3  |  Загрузок: 172

Открыть на HuggingFace →

Описание основано на материалах HuggingFace. Перевод выполнен автоматически.