[quantization] Complete StaticGemma4Runtime - #822

Draft

dvsav wants to merge 11 commits into

Samsung:mainfrom

dvsav:gemma_runtime

Commits on Jul 9, 2026

[quantization] Complete StaticGemma4Runtime
d.savchenkov
committed

Commits on Jul 10, 2026

fix missing image start token in the prompt
d.savchenkov
committed
fix missing image_position_ids
d.savchenkov
committed
fix missing PLE inputs
d.savchenkov
committed

Commits on Jul 13, 2026

fix head dimension mismatch
d.savchenkov
committed
fix tensor dimension mismatch in layer_caches
d.savchenkov
committed
fix missing shared-KV for shared-KV layers
d.savchenkov
committed
fix Wrong attribute path to access
d.savchenkov
committed
fix Shared-KV layers return only hidden_states
d.savchenkov
committed
fix Key length exceeds causal mask capacity
d.savchenkov
committed

Commits on Jul 15, 2026

build static inputs validation
d.savchenkov
committed