CYBER WATCH / EQUINOXE INTELLIGENCE

← All articles

b12237: mtmd: support Gemma 4 input audio stream (#31965)

llama.cpp Releases · 2026-11-10 04:01:09

The mtmd update adds support for Gemma 4 input audio streams, using a 480 ms step. The server enables real‑time embeddings only for audio generation. The real‑time slot forces embeddings on, marking every row of a batch as an output. Since only an audio‑generating model reserves n_batch outputs, a chat model with a Gemma 4 audio stream aborts on its first batch of audio rows. The hidden state is read only by the audio generation pipeline. Co‑authored by Pascal admin@serveurperso.com

Read original source ↗