software-mansion/react-native-executorch-gemma-4

1---2license: gemma3pipeline_tag: text-generation4tags:5- google6- deepmind7- gemma48- edge9base_model:10- google/gemma-4-E2B11library_name: executorch12---13 14# gemma-415 16This repository hosts the **gemma-4** models exported for the17[React Native ExecuTorch](https://www.npmjs.com/package/react-native-executorch)18library as ExecuTorch `.pte` programs, ready to run on device.19 20## Variants21 22| Path | Backend | Precision |23| --- | --- | --- |24| `e2b/mlx/gemma4_e2b_mlx_int4.pte` | mlx | int4 |25| `e2b/vulkan/gemma_4_e2b_vulkan_8da4w.pte` | vulkan | 8da4w |26| `e2b/xnnpack/gemma_4_e2b_xnnpack_8da4w.pte` | xnnpack | 8da4w |27 28## Repository structure29 30```31config.json                                28 B32e2b/mlx/config.json                        1.2 kB33e2b/mlx/gemma4_e2b_mlx_int4.pte            2.7 GB34e2b/tokenizer.json                         30.7 MB35e2b/tokenizer_config.json                  21.8 kB36e2b/vulkan/config.json                     1.3 kB37e2b/vulkan/gemma_4_e2b_vulkan_8da4w.pte    2.4 GB38e2b/xnnpack/config.json                    1.3 kB39e2b/xnnpack/gemma_4_e2b_xnnpack_8da4w.pte  2.5 GB40```41 42## Compatibility43 44These files are published for the **ExecuTorch v1.4.1** runtime. ExecuTorch45gives no forward compatibility guarantee, so an older runtime may fail to load46them.47 48To use them in React Native ExecuTorch, pass the model constant shipped in the49library's model registry to the corresponding task pipeline. See the50[documentation](https://docs.swmansion.com/react-native-executorch/docs/fundamentals/downloading-models).51 52To load these files in your own ExecuTorch runtime, read the53[compatibility note](https://github.com/pytorch/executorch/blob/main/runtime/COMPATIBILITY.md)54first.55