Add Qwen3-VL vision-language model support - #442

Open

nyo16 wants to merge 15 commits into

elixir-nx:mainfrom

nyo16:feat/qwen3-vl

Commits on Jan 6, 2026

feat: Add Qwen3-VL (Qwen2-VL) vision-language model support
nyo16
committed
b564150
fix: Correct parameter mapping for Qwen3-VL model loading
nyo16
committed
7ffb2c1
fix: Fix Qwen3VL featurizer argument order and image sizing
nyo16
committed
7596232
fix: Implement proper 2D rotary position embedding for Qwen3-VL vision encoder
nyo16
committed
8fc4038
Update Qwen3-VL config loader and test reference values
nyo16
committed
147da1f
Enable Qwen3-VL test with tiny model from HuggingFace
nyo16
committed
35479af
Remove Qwen2-VL mappings (not tested, different param naming)
nyo16
committed
c805c32
Add Qwen3-VL Livebook with examples and test documentation
nyo16
committed
b07ac6b

Commits on Jan 7, 2026

Remove appendix from Qwen3-VL Livebook
nyo16
committed
57553e7
Add DeepStack feature mergers for Qwen3-VL
nyo16
committed
b1e28f3
Implement full DeepStack injection for Qwen3-VL
nyo16
committed
8548ee3

Commits on May 24, 2026

Commits on May 27, 2026

Fix off-by-one in DeepStack layer indexing
nyo16
committed
9e87345