This is a full-expert GLM-5.2 hybrid checkpoint that combines the compact MXFP8/NVFP4/NF3 layout from madeby561/GLM-5.2-MXFP8-NVFP4-NF3-Hybrid with selected refusal-reduction tensors from Jiunsong/SuperGLM-5.2-abliterated-NVFP4. It retains all 256 routed experts per layer and is intended for custom Blackwell inference stacks that support the upstream hybrid format. The checkpoint is not affected, but the upstream serving runtime has a known topology hazard. The pinned July 26 v20 image assumes that all participating GPUs can use its B12X PCIe one-shot allreduce path. On four-GPU hosts where P2P works only within GPU pairs rather than across every rank, worker initialization can hang indefinitely while allocating shared CUDA IPC buffers. NCCL may already have initialized successfully, so the stall can look like a compile or model-load delay. If any off-diagonal rank pair in CUDAVISIBLEDEVICES reports NS, do not use the stock v20 GLM launcher unchanged. The affected pinned image is: The July 30 r13 image is not yet a safe replacement for partial-mesh topologies. Although its vLLM default is opt-in, the image environment still sets VLLMENABLEPCIEALLREDUCE=1, and the GLM launcher…
Модальности:
Генерация текста
Области применения:
Диалог / чат
Задача: Генерация текста
Автор: jacklarmer
Теги: glm_moe_dsa, glm, glm-5.2, mixture-of-experts, abliterated, mxfp8, nvfp4, nf3
Лайков: 4 | Загрузок: 173
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
