GGUF of GLM-5.2, REAP expert-pruned (50%) + Q2K (~129 GB) — the maximum-context option for 2× 96 GB GPUs (192 GB), leaving ~60 GB for KV cache (≈2.5× the context room of the Q3K build). This stacks REAP-50 (~+37.5% perplexity vs full GLM-5.2) with 2-bit Q2K. It works but is delicate: — ✅ Coherent with sampling: «The capital city of France … is Paris.» — ❌ Collapses into repetition ( ) with greedy / temp 0 decoding.** If you can spare the VRAM, the Q3KM build (~169 GB) is noticeably more robust. For real quality, use the MLX REAP-25 (+2.3% PPL) or full GLM-5.2. Stock llama.cpp can’t load GLM-5.2 GGUFs yet (DSA indexer required on every layer; GLM-5.2 ships it on only some). INVALID LANGUAGE PAIR SPECIFIED. EXAMPLE: LANGPAIR=EN|IT USING 2 LETTER ISO OR RFC3066 LIKE ZH-CN. ALMOST ALL LANGUAGES SUPPORTED BUT SOME MAY HAVE NO CONTENT
Модальности:
Генерация текста
Области применения:
Диалог / чат
Задача: Генерация текста
Автор: pipenetwork
Теги: gguf, llama.cpp, moe, glm, reap, pruned, endpoints_compatible, conversational
Лайков: 3 | Загрузок: 95
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
