Uncensored GGUF variant of AtomicChat/DeepSeek-V4.1-Flash-1017-Uncensored-GGUF, converted to ISTA-DASLab GSQ-RCO IQ1M and IQ2XXS shards with a shared Engram IQ4NL file. It is intended for one AMD Strix Halo machine (gfx1151`, 128 GB unified memory) using the Gufo fork: https://github.com/franzmoca/gufo/tree/feat/ds41 Usage notes: https://github.com/franzmoca/gufo/blob/feat%2Fds41/README_DEEPSEEK.md This is an uncensored model. Review the upstream model card and use it responsibly. Drafter credit: the DSpark drafter was contributed by Lucebox/DeepSeek-V4.1-Flash-1017-ROCMFP-GGUF. The 00006 shard of both IQ1M and IQ2XXS points to the same Engram IQ4NL` file. AMD Ryzen AI Max+ 395 (128 GB), AMD Radeon 8060S (gfx1151), Ubuntu 26.04 LTS. Memory note: this sharded model with streamed experts runs at extreme memory pressure on 128 GB. Inference works reliably only when no other GPU services are running (no other model servers, containers, or concurrent model loads). For DSpark speculative decoding, append the repo’s drafter (contributed by Lucebox/DeepSeek-V4.1-Flash-1017-ROCMFP-GGUF): These results were measured without the DSpark drafter. With the DSpark drafter, generation is about 30…
Модальности:
Генерация текста
Задача: Генерация текста
Автор: dineshr93
Теги: gguf, deepseek, uncensored, strix-halo, gufo
Лайков: 3 | Загрузок: 779
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.