software-mansion/react-native-executorch-whisper-tiny
1167k
1---2license: apache-2.03pipeline_tag: automatic-speech-recognition4library_name: executorch5---6 7# whisper-tiny8 9This repository hosts the **whisper-tiny** models exported for the10[React Native ExecuTorch](https://www.npmjs.com/package/react-native-executorch)11library as ExecuTorch `.pte` programs, ready to run on device.12 13Upstream model: [whisper-tiny](https://huggingface.co/openai/whisper-tiny)14 15## Variants16 17| Path | Backend | Precision |18| --- | --- | --- |19| `coreml/whisper_tiny_coreml_fp16.pte` | coreml | fp16 |20| `mlx/whisper_tiny_mlx_bf16.pte` | mlx | bf16 |21| `mlx/whisper_tiny_mlx_int8.pte` | mlx | int8 |22| `vulkan/whisper_tiny_vulkan_int8.pte` | vulkan | int8 |23| `vulkan/whisper_tiny_vulkan_fp16.pte` | vulkan | fp16 |24| `xnnpack/whisper_tiny_xnnpack_fp32.pte` | xnnpack | fp32 |25 26## Repository structure27 28```29config.json 34 B30coreml/config.json 1.5 kB31coreml/whisper_tiny_coreml_fp16.pte 72.9 MB32mlx/config.json 2.6 kB33mlx/whisper_tiny_mlx_bf16.pte 72.6 MB34mlx/whisper_tiny_mlx_int8.pte 57.1 MB35tokenizer.json 2.4 MB36vulkan/config.json 2.6 kB37vulkan/whisper_tiny_vulkan_fp16.pte 204 MB38vulkan/whisper_tiny_vulkan_int8.pte 124 MB39xnnpack/config.json 1.4 kB40xnnpack/whisper_tiny_xnnpack_fp32.pte 222 MB41```42 43## Compatibility44 45These files are published for the **ExecuTorch v1.4.1** runtime. ExecuTorch46gives no forward compatibility guarantee, so an older runtime may fail to load47them.48 49To use them in React Native ExecuTorch, pass the model constant shipped in the50library's model registry to the corresponding task pipeline. See the51[documentation](https://docs.swmansion.com/react-native-executorch/docs/fundamentals/downloading-models).52 53To load these files in your own ExecuTorch runtime, read the54[compatibility note](https://github.com/pytorch/executorch/blob/main/runtime/COMPATIBILITY.md)55first.56 