PulseAugur
EN
LIVE 06:57:39

llama.cpp b10438 release fixes Granite4 Vision image assembly

The llama.cpp project has released version b10438, which includes a fix for the Granite4 Vision image sequence assembly. This update addresses issues with truncation for scaled image height and width before unpadding, and also removes debug scaffolding related to MTMD_DUMP_EMBD. The release also involves cleanup of comments and removal of dead code. AI

IMPACT Minor update to a tool used for running LLMs, improving image sequence handling for Granite4 Vision.

RANK_REASON This is a minor release/patch for a software project that enables running LLMs, not a new model release or significant research.

Read on llama.cpp — Releases →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

llama.cpp b10438 release fixes Granite4 Vision image assembly

COVERAGE [1]

  1. llama.cpp — Releases TIER_1 English(EN) · hbattu73 ·

    b10438: mtmd: fix Granite4 Vision image sequence assembly (#26653)

    <ul> <li>mtmd: fix granite 4v grid assembly</li> </ul> <p>(cherry picked from commit <a class="commit-link" href="https://github.com/ggml-org/llama.cpp/commit/91f82eb1b489bff0dc3f649edf2cb9e0b7f656e9"><tt>91f82eb</tt></a>)</p> <ul> <li>mtmd: fix truncation for scaled image height…