gemma-4-19b-a4b-it-REAP-GGUF
GGUF quantizations of 0xSero/gemma-4-19b-a4b-it-REAP, a 30% expert-pruned variant of google/gemma-4-26b-a4b-it using the REAP (Router-weighted Expert Activation Pruning) method....
GGUF quantizations of 0xSero/gemma-4-19b-a4b-it-REAP, a 30% expert-pruned variant of google/gemma-4-26b-a4b-it using the REAP (Router-weighted Expert Activation Pruning) method....
> [!TIP] > Support this work → · X · GitHub · REAP paper · Cerebras REAP 20%...
INVALID LANGUAGE PAIR SPECIFIED. EXAMPLE: LANGPAIR=EN|IT USING 2 LETTER ISO OR RFC3066 LIKE ZH-CN. ALMOST ALL LANGUAGES SUPPORTED...
𓌳 REAP𓌳 the Experts: Why Pruning Prevails for One-Shot MoE Compression 📄 Paper • 💻 Code • 📝...
> [!TIP] > Support this work → · X · GitHub · REAP paper · Cerebras REAP 50%...
The smallest of cerebras GPT models with only 111M parameters instruction fine-tuned. The model has been evaluated with...
This repository contains custom APEX (Adaptive Precision for EXpert Models) -inspired quants, tuned specifically to the underlying mixture-of-experts...
REAP удаляет 30% экспертов MoE (38 из 128 на каждый уровень), сохраняя при этом поведение маршрутизации модели. Количество...
— Prime Intellect — Для спонсирования вычислений и создания INTELLECT-3 — Cerebras — Для методологии сокращения REAP —...
𓌳 REAP𓌳 Эксперты: почему обрезка преобладает при однократном сжатии MoE 📄 Бумага • 💻 Код • 📝 Блог...