PR #12402 Add Qwen2.5VL support

Add Qwen2.5VL support #12402

ngxson merged 20 commits into ggml-org:master from HimariO:qwen25-vl

github-actions added examples

github-actions added python

HimariO force pushed from bc199879 to 0664dba3 159 days ago

HimariO marked this pull request as ready for review 159 days ago

HimariO force pushed from 4fadb823 to b28ad7ec 155 days ago

HimariO force pushed from b28ad7ec to b27a8dcb 149 days ago

ggerganov requested a review from

ngxson 144 days ago

ngxson commented on 2025-04-19

implment vision model architecture, gguf convertor

d9f2d71b

handle window attention inputs

3d5198ee

add debug utils

69b39add

fix few incorrect tensor memory layout

444e47c0

move position id remap out of ggml to avoid int32 cuda operations

2eb32933

cleaning up

d1af4598

ignore transformers Qwen2_5_xxx type check

91fbdd78

remove not so often use `qwen2vl-cli` debug functions

2de5dc3a

remove commented-out code blocks

7f530ac0

fix attn weight scaling after rebase

a3cd0e52

ngxson commented on 2025-04-25

add `PROJECTOR_TYPE_QWEN2_5_VL`

caa7e57e

remove `KEY_USE_GLU_MLP`, `KEY_USE_RMS_NORM`

f69e9fa0

replace `KEY_FULLATTN_BLK_IDX` with `KEY_WIN_ATTN_PATTERN`

77b144a8

HimariO force pushed from b27a8dcb to 77b144a8 137 days ago

remove `attn_window_size` from gguf

7e1bb043

fix model conversion

516735ad

Merge branch 'master' into qwen25-vl

5085dbb2

clean up

0c74ea54

Merge branch 'master' into qwen25-vl

82f8e72e

fix merging problem

89be9199

add test

53a15d01

ngxson approved these changes on 2025-04-27

ngxson merged ca2bb89e into master 136 days ago

ngxson commented on 2025-04-27

Reviewers

ngxson

Assignees

No one assigned

Labels

examples python

Milestone

No milestone

llama.cpp Add Qwen2.5VL support #12402 Merged

Add Qwen2.5VL support #12402

llama.cpp
Add Qwen2.5VL support
#12402

Merged