llama.cpp b10369 发布：支持 pocket-tts 并优化性能

AI 摘要

llama.cpp 发布 b10369 版本，主要更新是支持 pocket-tts 文本到语音模型。该实现将转置卷积重构为 GEMM 加 col2im，使 CUDA 上每帧生成时间降低 80%，CPU 上降低 50%，输出与参考实现高度一致。同时新增了语言包特定参数（如帧尾填充、短文本填充）和模型变体支持，并更新了文档。

正文节选

<details open> mtmd: support pocket-tts (#26871) * adapt the api * text model ok * working impl, need verify and clean up * mtmd: build the pocket-tts transposed convolutions as GEMM + col2im ggml_conv_transpose_1d has no grouped mode, so the depthwise upsample was built as one convolution and one concat per channel, which floods the graph with small nodes and makes kernel launches dominate the decoder. Fold both cases into the column form the seanet decoder already needs: the general cas

发布时间：2026-08-12 12:52

抓取时间：2026-08-12 13:10

来源机构：ggml-org