Community GGUF quantizations of gnani/gnani-evon-v3.3-30B-A3B, developed by Gnani.ai. Credit for the model, its training, and its reported evaluations belongs to Gnani.ai and the upstream model authors. This repository contributes format conversion, quantization, and the checks documented below. The weights were not trained further for this release. This is not an official Gnani.ai or NVIDIA release. The motivation is appreciation for Gnani.ai’s work and a wish to make it easier to run. The original card lists «Supported Hardware: NVIDIA H100 80GB, NVIDIA H200, NVIDIA A100». These smaller files offer a route to running the model on higher-end consumer systems or affordable rented GPUs through llama.cpp, including partial GPU offload with the remaining weights in system RAM. Our short tests ran on NVIDIA A10 hardware with ten layers offloaded. Consumer GPUs and other runtimes have not been tested here, and memory requirements depend on the quant, context, and runtime settings. For the model’s capabilities, intended uses, training, and evaluations, please start with the original model card. Its results describe the original model. We have not reproduced those evaluations on these…
Модальности:
Генерация текста
Области применения:
Логика и рассуждение Диалог / чат
Задача: Генерация текста
Автор: luminenio
Теги: llama.cpp, gguf, nemotron_h, mamba, moe, indic, long-context, reasoning
Лайков: 3 | Загрузок: 880
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
