HuggingFace’s download widget does not recognize KS and KT ggufs ——> Text-only GGUF conversions of Qwen3.8-27B, quantized with an importance matrix calibrated for agentic coding rather than general text. Qwen3.8-27B ships as a vision-language model. These files carry the text tower and MTP block only. The vision tower is not converted and is not present, so these will not process images or video. Vision requires the original weights with a runtime that supports them. The quantization selector on this page lists only IQ3XXS, and the «Use this model» snippets are all pinned to it. That is a display limit, not a statement about the other files. The Hub builds that list from general.filetype, and it recognizes the values in upstream llama.cpp’s enum. IQ3XXS is 23 and is recognized. IQ4KS is 145 and IQ3KT is 152, both ikllama-specific and outside that enum, so neither appears. Use the links above, or the Files tab. tokenembd is not a layer, so -ngl 99 does not move it to the GPU. Host-side embedding stays in system memory rather than VRAM and no flag is needed. Its precision therefore costs VRAM nothing but adds download size. IQ3KT carries it at q8_0; the other two carry it at the…
Модальности:
Генерация текста
Области применения:
Диалог / чат
Задача: Генерация текста
Автор: ji-farthing
Теги: gguf, ik_llama, qwen35, imatrix, text-only, endpoints_compatible, conversational
Лайков: 4 | Загрузок: 2,579
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
