Gemma4 model implementation - #460

Draft

edenamram wants to merge 7 commits into

elixir-nx:mainfrom

edenamram:gemma4-model-implementation

Commits on Jun 9, 2026

Add Gemma4 model implementation and update model mappings
edenamram
committed
ba6d493

Commits on Jun 13, 2026

Gemma4Text to enhance cache initialization and attention head size handling
edenamram
committed
2c21e11
Implement per-layer input handling in Gemma4 model and add ChunkedFileTensor for large tensor support
edenamram
committed
41a2d03

Commits on Jun 15, 2026

Commits on Jun 16, 2026

Refactor rotary embedding options and improve parameter mapping in Gemma4Text
edenamram
committed
0d707a6

Commits on Jun 18, 2026

Refactor Gemma4Text to remove shift parameter from normalization layers and enhance attention block implementation
edenamram
committed
425111c