]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
mtmd : fix Pixtral OOM with large images by capping image_size to 1024 (#14326)
authoryuiseki <redacted>
Sun, 22 Jun 2025 12:44:57 +0000 (21:44 +0900)
committerGitHub <redacted>
Sun, 22 Jun 2025 12:44:57 +0000 (14:44 +0200)
commit5d5c066de8a3d2cb32f04c4d5ad1560945f30bf3
tree9563ef5af12862421e5676c3aa6581e24bb53a8a
parent40bfa04c95c19fb42bafd4e21b5c2a7771846801
mtmd : fix Pixtral OOM with large images by capping image_size to 1024 (#14326)

Mistral Small 2506 models using Pixtral vision encoder were running out
of GPU memory when processing images larger than 1024x1024 pixels due to
exponential memory growth from unlimited image size.

This fix applies the same 1024x1024 limit used by Qwen2VL models to
prevent OOM issues while maintaining compatibility with existing models.
tools/mtmd/clip.cpp