Repository navigation

cua_s1: complete native vision and screenshot inference with GPU parity - #64

Merged

hsliuustc0106 merged 24 commits into

ThinkFlowLab:mainfrom

Levius-Fubuki:codex/cua-native-vision

Oct 6, 2026

Commits on Oct 1, 2026

cua_s1: accept adapted image rows and three-axis language positions
Levius-Fubuki
committed
75a80cb
cua_s1: load and validate native vision checkpoints on CPU
Levius-Fubuki
committed
de46f07

Commits on Oct 2, 2026

cua_s1: preprocess RGB images for native vision on CPU
Levius-Fubuki
committed
591097e
Merge branch 'codex/cua-vision-loader' into codex/cua-native-vision
Levius-Fubuki
committed
6844de9
Merge branch 'codex/cua-image-preprocess' into codex/cua-native-vision
Levius-Fubuki
committed
35c16e0
Merge branch 'codex/cua-native-multimodal-input' into codex/cua-native-vision
Levius-Fubuki
committed
6c40bba
feat(cua_s1): execute native vision and serve screenshot decisions
Levius-Fubuki
committed
932e645
chore(cua_s1): limit PR diff to core implementation
Levius-Fubuki
committed
f73ce6a
chore(cua_s1): limit PR diff to core implementation
Levius-Fubuki
committed
1512852
chore(cua_s1): limit PR diff to core implementation
Levius-Fubuki
committed
63f974f
chore(cua_s1): limit PR diff to core implementation
Levius-Fubuki
committed
9ee058f

Commits on Oct 5, 2026

Commits on Oct 6, 2026