llama.cpp
mtmd : support Qwen 2.5 Omni (input audio+vision, no audio output)
#13784
Merged
Merged
mtmd : support Qwen 2.5 Omni (input audio+vision, no audio output) #13784
ngxson
merged
17 commits into
ggml-org:master
from
ngxson:xsn/qwen25omni
mtmd : allow multiple modalities at the same time
58c7849f
refactor mtmd tokenizer
2782a583
fix compile
bb92d1d0
ok, missing SinusoidsPositionEmbedding
8b51e7fa
first working version
24ec43eb
fix style
1ac73f40
more strict validate of n_embd
90132453
refactor if..else to switch
346d2522
fix regression
6e65e0c5
add test for 3B
235fbdbf
update docs
bf34f38f
fix tokenizing with add_special
d03c2407
add more tests
ef48e8f2
fix test case "huge"
94d893d6
Merge branch 'master' into xsn/qwen25omni
baa882a5
rm redundant code
05310968
set_position_mrope_1d rm n_tokens
27a8f266
Reviewers
Assignees
No one assigned
Labels
documentation
examples
python
Milestone
No milestone
Login to write a write a comment.
Login via GitHub