llama.cpp v0.9.0 发布：FP8 量化与多序列 KV 优化
原标题：v0.9.0
AI 摘要
llama.cpp 发布 v0.9.0，新增 OCP FP8 E4M3 量化支持、提升多序列推理统一 KV 缓存效率的 flash_attn_ext_rows 算子，以及通用批量张量拷贝 API。该版本新增 Clover、Limite、Metis-1.6、Mistral Large 4 四种架构支持，扩展 Chatterbox TTS、Nemotron 3.5 Transcribe 等音频能力，并对 Gemma 4 E2B/E4B 做大幅 prefill 优化。底层 ggml 升级至 v0.29.0，带来 NCCL 多节点张量并行及多后端性能改进。
正文节选
## Overview This release adds OCP FP8 E4M3 quantization support, a new `flash_attn_ext_rows` op that makes the unified KV cache efficient for multi-sequence inference, and a generic batched tensor copy API. Four new architectures are supported (Clover, Limite, Metis-1.6 and Mistral Large 4), audio generation and transcription expand with Chatterbox TTS, Nemotron 3.5 Transcribe and predefined Pocket TTS voices, and the Gemma 4 E2B/E4B models get major prefill optimizations. `ggml` is updated t