Nanbeige4.2-3B on Apple Silicon: Fixing Deployment Bugs and Decreasing Looped Transformer Memory Overhead

• 类型：arxiv
• 标识：2608.13987
• 链接：https://arxiv.org/abs/2608.13987
• 主分类：agent
• 形态：application
• 被引：0
• 被引来源：Semantic Scholar
• S2被引：0
• 影响力被引：0
• TLDR：A chunked-prefill strategy is introduced which alleviates the incurred memory-capacity penalty, extending allowable context width by $2.7 \times$ on 32~GiB shared memory, however, even with the reduced memory overhead, it is shown that patches are required to render Nanbeige4.2-3B usable.
• 副分类：engineering
• 待LLM分类：否
• 标题中文：Nanbeige4.2-3B on Apple Silicon：修复部署 Bug 并降低 Looped Transformer 显存开销
• TLDR中文：提出一种 chunked-prefill 策略，可缓解由此带来的内存容量惩罚，在 32 GiB 共享内存上将允许的上下文宽度扩展 $2.7 \times$，但即便降低了内存开销，仍需打补丁才能使 Nanbeige4.2-3B 可用。
• 来源文件：
• /inbox/tom/_candidates/2026-08-18-rag-retrieval-reranking-candidates.json
• /inbox/tom/_candidates/2026-08-18-agent-rag-longcontext-candidates.json
• [S2 enrich]