nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 17 Packages Projects Releases Wiki Activity

mtmd: support "frame merge" for qwen-vl-based models (#21858)

Browse Source 
* feat: add video support for Qwen3.5

* various clean up

* revise the design

* fix llava-uhd case

* nits

* nits 2

---------

Co-authored-by: andrewmd5 <1297077+andrewmd5@users.noreply.github.com>
This commit is contained in:
Xuan-Son Nguyen
2026-06-06 21:17:25 +02:00
committed by GitHub
co-authored by andrewmd5
parent 6b80c74f28
commit 31e82494c0
10 changed files with 197 additions and 74 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tools/mtmd/mtmd-image.cpp
+1 -1
View File
@@ -1116,7 +1116,7 @@ bool mtmd_image_preprocessor_deepseekocr::preprocess(const clip_image_u8 & img,
static constexpr int native_resolutions[] = { 1024 /* base */, 1280 /* large */ };
// TODO: support 512 (tiny) and 640 (small) once we have eval data for them
const int64_t orig_area = static_cast<int64_t>(img.n_pixels());
const int64_t orig_area = static_cast<int64_t>(img.get_size().area());
size_t mode_i = 0;
int64_t min_diff = std::numeric_limits<int64_t>::max();